Thales-Bristol Partnership in Hybrid Autonomous Systems Engineering (T-B PHASE)

Lead Research Organisation: University of Bristol
Department Name: Computer Science

Abstract

Hybrid autonomous systems are those where groups of people are in direct, ongoing interaction with groups of autonomous robots or autonomous software.

One prominent current example involves rush-hour traffic made up of a mixture of cars driven by people and cars driven by smart algorithms. However, emerging technologies in robotics, AI and ICT mean that hybrid autonomous systems of this kind will become increasingly common in a much wider set of situations:

Emerging technologies in robotics, AI and ICT mean that hybrid autonomous systems of this kind will become increasingly common in a much wider set of situations:

- a mixture of autonomous and human-operated drones making deliveries or monitoring public spaces;
- a mixture of human traders and autonomous trading agents buying and selling stocks;
- a mixture of autonomous and human-operated trains and trams providing efficient, integrated public transport;
- autonomous systems assisting with search and rescue missions in disaster areas that are difficult or dangerous to access;
- robot carers assisting care workers with the provision of social care in the home

In each of these cases smooth, reliable, safe interaction amongst machines and people will be key to success. But how can we guarantee that self-driving cars won't cause a crash or gridlock? How can we understand how autonomous systems will respond to new situations (both acute shocks and long-term gradual changes in their environment), or changes in the way that people interact with them? Consequently, as we enter this new design space, a crucial challenge for the engineers of hybrid autonomous systems across all of these settings is ensuring that the system behaviour is Robust and Resilient and that it meets Regulatory demands: the R3 Challenge.

T-B PHASE directly addresses this R3 Challenge for Hybrid Autonomous Systems Engineering, by bringing together expertise in robotics, AI, and systems engineering at the University of Bristol and Thales in a five-year project that targets fundamental autonomous system design problems in the context of three real-world Thales use cases: Hybrid Low-Level Flight, Hybrid Rail Systems, and Hybrid Search & Rescue.

Bristol and Thales have a long-standing track record of research collaboration, and by jointly pursuing fundamental research questions in the context of highly practical design problems, alongside a programme of engagement with industry, the public and regulatory bodies, T-B PHASE will significantly advance our capability to operate confidently in one of the most important emerging areas for modern engineering.

Planned Impact

The business benefits of T-B PHASE are:

- Liability and responsibility for the behaviour of complex systems strongly inhibits the deployment of hybrid autonomous systems in the real world. By embedding Robustness, Resilience and Regulation as part of the development life cycle, T-B PHASE will provide those who commission and develop hybrid autonomous systems with tools that enable early-stage evaluation and demonstration in the development lifecycle.

- T-B PHASE will accelerate the adoption of new hybrid autonomous systems by reducing the costs of development, the risks of deployment, and the length of the development life-cycle. Current approaches involve exhaustive real-world validation and verification campaigns that are not scalable or sustainable for systems with emergent properties.

- Public and consumer acceptance of autonomous hybrid systems is currently fragile. T-B PHASE will improve the transparency of system Robustness and Resilience, which is an essential aspect of building acceptance and evolving Regulation frameworks that are suitable for hybrid autonomous systems.

- Developing a critical mass of skilled researchers in autonomous systems engineering will benefit multiple relevant sectors in the UK, stimulating further technology development in this area and creating longer term strategic benefits for the UK engineering sector.

- T-B PHASE will contribute to advancing the development of regulatory frameworks for autonomous systems, which require attention to what can be assured at design stage and what can be assured post-deployment through online monitoring and adaptation.

The pathways to achieving this impact are described fully in the Pathways to Impact section of this proposal.
 
Description At the end of year 2 of T-B PHASE the project has yielded the following highlighted research, for which the associated outputs have been reported under relevant sections of this report:

Theme 1: Self-Monitoring in Context (Pitonakova, University of Bristol):
An autonomous robot's ability to detect novelty in its own performance or in its environment could make it more robust and safer. This is particularly important given that such a robot may operate in an open environment that changes over time and may itself change due to learning, adaptation, degradation, failure or malicious attack. This study considered using Grow When Required Neural Networks (GWRNNs) as a way of learning to detect novelty in a robot's environment.

First, we have shown that there are two important parameters affecting GWRNN performance and learning speed -- the number of neuron input connections (which determines how many features a network can consider at a given point in time), and the neuron activation threshold (which controls the network's growth). We have demonstrated that using a new Plastic variant of the GWRNN, where the value of the first parameter varies from one clustering neuron to another based on the size of input vectors that they represent, leads to more robust performance. In general, we conclude that it is desirable to create adaptation mechanisms that automatically adjust a novelty detector's parameters but that are parameter-free themselves.

Secondly, we have demonstrated that a novelty detector may perform fundamentally differently in the one-shot and in the continuous variants of the novelty detection task. In the latter, a series of successive distinct novelties take place sequentially and a novelty detector needs to be able to distinguish between different novelties that may have similar features but that occur in different places or at different times. This may be especially important in robotic applications such as surveillance and intruder detection.

Thirdly, we have also shown that performance of a novelty detector may vary significantly when it comes to recognising different types of novel events. For instance, the ability to notice the appearance of previously unseen objects is fundamentally different from the challenge of noticing the disappearance of familiar objects.

Finally, we have demonstrated that adding localisation information to the sensory input vector of a novelty detector often improves its ability to distinguish between different objects and to detect the absence of previously learned objects. It should be noted, however, that the impact of location data noise on learning speed and on the likelihood of reporting false positives needs to be studied in detail.

Themes 2: Consensus Formation for Collaborative Autonomy; and Theme 3: Dynamical Hierarchical Task Decomposition and Planning for Heterogeneous Systems (Crosscombe and Kent, University of Bristol):
There are many real-world problems that can be effectively solved using multiple agents, such as search and rescue, surveillance, monitoring and mapping. This can result in a need for coordination, cooperation and ultimately communication that may or may not be possible in many situations. Different problems may also dictate to what extent we require the ability to distribute and/or decentralise our decision making.

Many off-the-shelf solutions, such as UAVs, are designed in isolation and typically focus on single-agent behaviours. Our research question concerned how we could plan for and deploy autonomous solutions which could appropriately bridge the gap between single and multi-agent behaviours, when individuals may possess differing objectives, constraints or beliefs about the state of the world. Importantly, how could the levels of de-centrality in the decision making and communication affect the way in which we design and deploy these autonomous solutions?

We developed several frameworks for distributed learning in which individual agents, through local interactions, could reach a consensus about the true state of the environment they inhabited. This is an important feature for systems that need to coordinate and to act in a collective way. In these models, agents learned more efficiently by communicating with their neighbours and sharing the information they had gathered, and furthermore the process by which information was fused was shown to be more robust to noise than a system in which agents learned individually. The resulting models produced systems of agents which made better decisions because they based their decisions on more accurate information.

The main outcome of this result is that we have improved our understanding of how decentralised systems provide increased robustness when compared with centralised systems. This is typically assumed but not currently well understood or studied. For example, the most efficient deployment of multiple autonomous submarines for the task of ocean surveillance would likely be a partitioned mapping of agents to segments of the ocean to be explored. However, while this is certainly an efficient deployment, such an approach cannot counter the effects of noise in the system, e.g., sensor noise. If, for example, a sensor interprets an input signal incorrectly at a rate of 20%, then the average error of the systems will be around 20%. An alternative is to deploy a fully decentralised system in which the submarines share the same boundaries and can revisit regions of the ocean that would normally be allocated to a single submarine. The result is a system which corrects for noise and achieves close to 100% accuracy on average.
Alongside this work we have also demonstrated, for the multi-agent persistent surveillance problem, that if your system consists of any homogenous agents then some degree of noise could in fact be a desirable property. This is a reassuring balance, as in many real-world scenarios noise is an unavoidable component of your system, through imperfections in aspects such as sensing, communication or computation. In fact, in this work we outlined how properties of a standard state, policy, action decision cycle could lead to highly undesirable emergent multi-agent behaviour.

Theme 4: Cascading Failure and Network Topology (Rayneau-Kirkhope, Thales):
Research led by a Thales researcher on Cascading Failure undertook to understand and develop techniques for predicting the loading of the UK Rail Transport Network under both failure and movement models based on current, future legacy and future train movement rules. Additionally, the research looked at the demand for resources (trains) against loading with changing constructs of operation. This created knowledge around stress-point localisation, optimised loading capability and asset planning that may be utilised for future network planning and for on-line network re-planning in the event of disruptions and failure.

Benefits are predicted in the areas of network and fleet optimisation for service and the prediction of failure effects and propagation. This should benefit the rail passenger, rail operating companies and the rail infrastructure / track operations.

This work has led to the investigation of in-house proof-of-concept and technical demonstrators options at Thales.
Exploitation Route We are working closely with Thales Global Business Units, use case experts and stakeholders to focus the research in a direction that could potentially be exploited within the company.

Previous Thales experience indicates that the most effective and beneficial method of working is where academic researchers are embedded into the design teams for a period of time and a relationship between the parties develops to allow a freer exchange of views and ideas. This gives the researchers the most effective views of the working practices and also allows the design teams to benefit directly from the interaction and potential adoption of new techniques.

Method development may form some of the research areas and the experimental application in the above manner using embedding has been seen to have benefit. Potentially, the embedding of techniques in Thales tools for evaluation is also seen as constructive prior to adoption / modification. T-B PHASE uses a modified version of the foregoing where the academic and Thales researchers are co-located and where the team will be embedded with design teams for a period of time. This is intended to provide a freer exchange of views and ideas in a cohesive environment and is in contrast to contracted research where the academic task is simple sub-contracted as a rigid set of requirements.
Sectors Aerospace, Defence and Marine,Digital/Communication/Information Technologies (including Software),Transport

 
Title Autonomy and Resilience Working Method (D. Harvey) 
Description An initial survey activity has been carried out to assess how traditional systems engineering approaches can be applied to engineer resilient hybrid autonomous systems. This has included looking at general systems engineering approaches, resilience engineering and autonomy and subsequently identifying where significant changes may be needed. The rationale for this is to ensure that a 'top-down' analysis is carried out in order to identify pertinent research topics. An example of this is the approach to the verification and validation of autonomous systems; this will demand new approaches since conventional methods are ill-equipped to deal with the 'state space explosion problem' that is associated with autonomous systems that have the capacity to learn from their environment. 
Type Of Material Improvements to research infrastructure 
Year Produced 2018 
Provided To Others? No  
Impact Inform T-B PHASE project development and approaches. 
 
Title A Decentralised Multi-Demic Evolutionary Approach for solving Dynamic Multi-Agent Travelling Salesman Problems (Tom Kent) 
Description The Travelling Salesman and its variations are some of the most well known NP hard optimisation problems. This work looks to use both centralised and decentralised implementations of Evolutionary Algorithms (EA) to solve a dynamic variant of the Multi-Agent Travelling Salesman Problem (MATSP). The problem is dynamic as it is both carried out in simulation and solved at the same time, additionally during the simulation new tasks are added and completed tasks are removed. The problem is allocating an active set of tasks to a set of agents whilst simultaneously planning the route for each agent. 
Type Of Material Computer model/algorithm 
Year Produced 2018 
Provided To Others? Yes  
Impact The model and implementation have been used to research the impact of agent to agent communication distances on performance. The results and outline of this work are currently under review for publication. 
URL https://github.com/T-BPhase/MDEA
 
Title A multi-agent simulation environment to study consensus formation where beliefs are represented using Dempster-Shafer theory (Michael Crosscombe) 
Description In this model we study evidential updating in noisy environments where agents receive two kinds of evidence: direct evidence, received by exploring their environment, and indirect evidence, through consensus formation between pairs of agents. We have developed a general and extendable model for studying the various dynamics of this kind of system, and we have been exploring both symmetric and asymmetric updating methods. More specifically, we have studied two operators so far, Dempster's rule of combination and Dubois & Prade's operator, but additional operators can be added and studied. We have also explored alternative methods of updating beliefs based on direct evidence, such as negative updating 
Type Of Material Computer model/algorithm 
Year Produced 2018 
Provided To Others? Yes  
Impact Used in the development of new models within this project, studying the dynamics of different operators used in consensus formation, and to produce results suitable for publication. 
URL https://github.com/T-BPhase/dempster-shafer-consensus
 
Title ARGoS simulation implementation of the Grow-When-Required Neural Network for novelty detection on a robot (Lenka Pitonakova) 
Description The Grow-When-Required Neural Network is capable of learning an input space representation in a self-organised, unsupervised fashion and detect when a novel input is presented. The network runs inside a robot that travels through the environment, progressively learns its features and recognises when a novel object is placed into the environment. The simulation code varies the network, robot and environmental parameters and enables precise control of the variability of the environment and of the noise in the robot sensors. Data from the experiments is stored in text files with tab-separated values, organised in folders specified for each experiment. 
Type Of Material Computer model/algorithm 
Year Produced 2018 
Provided To Others? Yes  
Impact This model was used to produce results that were reported in "Robustness and Structural Plasticity of Grow-When-Required Neural Networks" submitted to the 2019 International Conference on Robotics and Automation. The page where the code is published was viewed by approximately 170 unique viewers since September 2018. 
URL http://lenkaspace.net/code/simulationModels/growWhenRequiredNeuralNetwork
 
Title Code and data set for "The robustness-fidelity trade-off in Grow When Required neural networks performing continuous novelty detection" (Lenka Pitonakova) 
Description Contains the simulation and analysis code and the entire data set 
Type Of Material Database/Collection of data 
Year Produced 2019 
Provided To Others? Yes  
Impact An open-source repository with simulation data and simulation and data analysis code was set up to accompany the paper published by Pitonakova and Bullock (2019) [1], in accordance with EPSRC's policy framework on research data. [1] Pitonakova L, Bullock S. (2019). The robustness-fidelity trade-off in Grow When Required neural networks performing continuous novelty detection. Neural networks : the official journal of the International Neural Network Society, 122, pp. 183-195. doi: 10.1016/j.neunet.2019.10.015 
 
Title Evolving behaviour trees to supervise simulated swarms (Elliott Hogg) 
Description Work initially carried out in a smaller research project in the MSc year has been extended during the first year of this PhD research. The work has provided proof-of-concept that a behaviour tree can be evolved by an algorithm to provide good supervision of a swarm, matching human-crafted heuristics for the same task. This work was expanded upon to explore an increased difficulty level which involved introduction of walls into the simulation environment and the study of the emergent behaviours of the swarm was reported in a paper presented at the 3rd International Symposium on Swarm Behaviour and Bio-Inspired Robotics (SWARM 2019) in November 2019 (reported as an output under publications, DOI to be added in 2020). Higher complexity scenarios will be applied and human trials will be conducted in the coming year of the PhD to extend this work. 
Type Of Material Computer model/algorithm 
Year Produced 2019 
Provided To Others? No  
Impact This work was presented at the 3rd International Symposium on Swarm Behaviour and Bio-Inspired Robotics (Swarm 2019) in November 2019 (reported as an output under publications). The simulation framework that has been designed and tested will enable us to begin thinking about how we could redesign the human-machine interface based on the trends we have observed. This leads on to planned human trials in the next year with the basis to perform meaningful studies that can be used to evaluate the benefits of the evolutionary framework. 
 
Title Networked-based state-of-the-world (SOTW) model (Michael Crosscombe) 
Description A networked-based state-of-the-world (SOTW) model was created within which agents attempt to explore and learn about their environment. A basic proof-of-concept network of one-hub-many-nodes (star network) was developed within it. This will enable experimentation with the most simplistic form of a network for the Thales Use-Cases and provide a comparison with a fully decentralised approach system. 
Type Of Material Computer model/algorithm 
Year Produced 2019 
Provided To Others? No  
Impact Anticipated for future publication under T-B PHASE. 
 
Title Preference-based simulation environment (Michael Crosscombe) 
Description The preference-based simulation environment (model) explores preference-based beliefs and is being extended to the same network-based communication structure as the networked-based state-of-the-world (SOTW) reported separately, to further our understanding of how simple belief models are affected by restrictions placed on agent communications. 
Type Of Material Computer model/algorithm 
Year Produced 2019 
Provided To Others? No  
Impact Anticipated for future publication under this project. Will inform a separate MSc project at the University of Bristol. 
 
Title Single agent policies for the multi-agent persistent surveillance problem (Tom Kent) 
Description An extension to the multi-agent simulator previously reported was developed incorporating an openAI-style 'gym' environment wrapper, allowing a common interface to test off the shelf machine-learning algorithms. This was used to train single agents to solve the persistent surveillance problem and be deployed in the multi-agent setting. 
Type Of Material Computer model/algorithm 
Year Produced 2019 
Provided To Others? No  
Impact Under review for publication in 2020. Lessons learned to inform overall T-B PHASE output anticipate for year 5. 
 
Title Traffic management simulator (William Bonnell) 
Description A robust and easy to use simulator was developed in the first year of this PhD research study. It aims to conduct traffic management experiments and builds on a simulator developed during the first year (MSc) project. A modular framework was implemented to allowing a range of UAV mechanics and high level traffic management methods to be applied to the simulation environment. Conflict avoidance was settled on as the standard approach, to ensure that any two UAVs are positioned at least a minimum desired distance from one another. Two basic performance metrics were developed; (1) flight time of a UAV and (2) the amount of time a UAV spends in conflict with other UAVs; providing a metric for efficiency and safety. Onward work on this will develop new metrics in order to explain and predict traffic performance, including introduction of waypoint routing and obstacle avoidance. These will form the basis of different traffic management methods to be developed and applied to the UAV crossroad scenario. 
Type Of Material Computer model/algorithm 
Year Produced 2019 
Provided To Others? No  
Impact The work will be developed throughout the remainder of the PhD with the aim to communicate the outcome via oral and written publications as appropriate. 
 
Description Thales/University of Bristol Collaboration under the T-B PHASE Prosperity Partnership grant 
Organisation Thales Group
Department Thales UK Limited
Country United Kingdom 
Sector Private 
PI Contribution Central to the T-B PHASE project is collaboration between Thales and Bristol researchers, supported by the co-location of academic and industrial of team members at the University.
Collaborator Contribution - Management and technical leadership to the research programme in partnership with the academic team at the University. - Project scoping and review. - Provision of Thales use cases for investigation to steer the programme. - Evaluation of research activity and review of applicability to use cases. - Engagement with and provision of relevant industrial stakeholders and customers (primarily Thales UK's Country Business Units and Use Case Domain Experts).
Impact [Note: To avoid duplication, specific research outcomes of this collaboration are reported under their respective headings for this award (e.g. papers, presentations, awards, event participation, engagement activities, research tools, methods, databases and models, placements/secondments and next destination of researchers).] Since the start of T-B PHASE, we have worked with Thales use case owners (technical specialists) in order to ensure our research is relevant to one or more areas of interest for the business. In Year 1 face-to-face meetings were held with Thales domain experts to determine specific requirements across the three main use cases that had been identified at the start of the programme. These discussions helped to steer the research by creating 'toy problems' that could be considered by the research team in the development of their research. In addition, two Use Case Workshops with Thales domain experts, stakeholders and internal customers were held during Year 1, exploring a range of issues and relating these back to the initial research direction of T-B PHASE. In Year 2 discussions were held with the Thales Information, Surveillance and Reconnaissance (ISR) team, enabling T-B PHASE researchers to identify how the algorithms they are developing could interface with existing Thales simulators, such as those that are designed for autonomous sensor management. Also in Year 2, members of the research team undertook site visits to rail operative control rooms in Cardiff and London, and held direct discussions with the Thales Ground Transportation System (GTS) teams. These activities informed the research theme on Cascading Failure and Network Topology and will be further developed under Hybrid Autonomous Systems (HAS) theme, which will focus primarily on human factor influences on HAS systems. More generally, throughout the duration of the T-B PHASE programme the team have produced internal reports that include workplans, project summaries and technical reports for each research theme in close collaboration with key Thales stakeholders who have reviewed and inputted to these. The team have participated in quarterly research and management review meetings with senior stakeholders from the Thales Systems Key Technical Domain (KTD) division to ensure the research progress and direction aligns closely to that required within the business. Future plans: Exploration of Thales stakeholder engagement beyond the UK is being explored, with the possibility of reaching Thales Germany and Canada through the human factors work on the GTS use case.
Start Year 2017
 
Description 18th Annual Conference on Systems Engineering Research (CSER) Conference (03/04/19) 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Supporters
Results and Impact The Conference on Systems Engineering Research, CSER, is the premier conference for systems engineering research in the world. It focusses on new, mostly theoretical, research on systems engineering and its applications. Ben Rayneau-Kirkhope (Thales Researcher on T-B PHASE) attended sessions which focussed on the systems engineering approaches to complex systems as relevant to T-B PHASE, i.e. those typically involving autonomous systems and AI. He networked with people working across industry using a similar approach to his research and applying this to real use cases, which could lead onto future discussion and possible engagement and ideas/data sharing.
Ben also participated in a Systems Engineering and Architecting doctoral networking event, SEANET prior to the main conference as an opportunity to find out what a number of PhD candidates were researching, some of which aligned to T-B PHASE research interests.
Year(s) Of Engagement Activity 2019
URL https://cser.info/
 
Description 1st T-B PHASE Advisory Group Meeting (15/05/2018) 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Other audiences
Results and Impact The 1st T-B PHASE Advisory Group meeting was hosted at University of Bristol on 15th May 2018. It involved nine individuals with experience of large-scale/enterprise level solutions engineering and/or academic research activity relevant to the aims of the T-B PHASE Research Project (including an EPSRC representative), as well as the academic and industrial project team members.

The Advisory Group were introduced to the T-B PHASE project and agreed the terms of reference for future Advisory Group sessions. Opportunity was provided for Advisory Group members to meet with T-B PHASE Postdoctoral Researchers and PhD students and to discuss their areas of expertise within the project directly.

The Advisory Group provided feedback and advice on the initial aims and objectives of the project and contributed to a research landscaping exercise to further define activities of relevance to the T-B PHASE scope (e.g. activities of interest within research, development and policy areas, and emerging research questions and initiatives).
Year(s) Of Engagement Activity 2018
 
Description 1st T-B PHASE Use Case Workshop (30-31/05/18) 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Industry/Business
Results and Impact 25 research team members and industrial stakeholders from University of Bristol and Thales Group attended the first T-B PHASE Use Case workshop in Bristol. The event facilitated detailed discussions on the T-B PHASE research challenges in relation to specified Use Cases within Thales, and included discussion on potential solutions/approaches or research areas to be followed with respect to each Use Case. The outcome of the event aims to inform the research direction of the T-B PHASE programme. The Research Team are currently developing research strands informed by the discussions.
Year(s) Of Engagement Activity 2018
 
Description 2nd T-B PHASE Advisory Group Meeting (29/11/2018) 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Other audiences
Results and Impact The 2nd T-B PHASE Advisory Group meeting was hosted at University of Bristol on 29th November 2018. It involved individuals with experience of large-scale/enterprise level solutions engineering and/or academic research activity relevant to the aims of the T-B PHASE Research Project (including an EPSRC representative), as well as the academic and industrial project team members. Membership has grown since the 1st meeting to 11 external members.
T-B PHASE researchers presented their current research activity to the group and there was a Q&A associated with this led by the Advisory Group Chair.
Feedback was provided on the current direction of the project and future plans for aspects of the project as yet not explored.
Finally, the Advisory Group discussed and advised on ways in which the T-B PHASE's low TRL academic research could best achieve a direct positive impact on its industrial stakeholders.
Year(s) Of Engagement Activity 2018
 
Description 2nd T-B PHASE Use Case Workshop (13-14/11/18) 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Industry/Business
Results and Impact 30 research team members and industrial stakeholders from University of Bristol and Thales Group attended the second T-B PHASE Use Case workshop in Thales UK HQ, Reading on 13-14 November 2018. The workshop aimed to identify how Thales could best utilise the outputs of the research programme and ensure these were of best value for the Business.
Ideas around achieving impact in Thales were discussed, and several areas of focus were identified. Routes to impact were reviewed and key delivery methods identified. There was specific focus on the opportunities presented for integration of autonomy into the Thales Business and planning was initiated for how to carry these forward.
Year(s) Of Engagement Activity 2018
 
Description 3rd T-B PHASE Advisory Group Meeting (12/12/19) 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Other audiences
Results and Impact The 3rd T-B PHASE Advisory Group meeting was hosted at University of Bristol on 12th December 2019 involving industrial and academic experts in hybrid autonomous systems engineering. A summary of the programme's status was given by Professor Eddie Wilson and Angus Johnson and feedback was given on several topics of interest, including Human-in-the-Loop integration (Human-Machine-Teaming) and the desired project outputs for T-B PHASE as a whole. Ethical issues relating to autonomous systems were also explored through an interactive workshopping activity, with links to relevant use cases made.

The recommendations of the Advisory Group following this meeting will influence decision-making processes with regards to the University and Thales' commitments to final project outputs.

In addition, ethical considerations for the research programme had had limited exploration within the workstreams and onward work would be influenced by the issues raised by the Advisory Group members.
Year(s) Of Engagement Activity 2019
 
Description Blog post: Fast Data Analysis Using C and Python 
Form Of Engagement Activity Engagement focused website, blog or social media channel
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Other audiences
Results and Impact Lenka Pitonakova wrote a blog post about a programming approach that increases data analysis code efficiency suitable for a non-scientific audience.
Year(s) Of Engagement Activity 2018
URL http://lenkaspace.net/code/dataScience/fastDataAnalysisCppPython
 
Description Blog post: Novelty Detection with Robots Using the Grow-When-Required Neural Network 
Form Of Engagement Activity Engagement focused website, blog or social media channel
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Other audiences
Results and Impact Lenka Pitonakova wrote a blog post about new research suitable for a non-scientific audience
Year(s) Of Engagement Activity 2018
URL http://lenkaspace.net/code/simulationModels/growWhenRequiredNeuralNetwork
 
Description EPSRC Council visit to Thales UK Headquarters (10/07/19) 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Other audiences
Results and Impact The T-B PHASE research programme was presented to the EPSRC Council as part of their visit to Thales in Green Park in July 2019. The academic and industry leads presented to the Council, with focus on the benefits of a partnership working approach and suggested areas for improvement in the overall prosperity partnership model. A dedicated stand of T-B PHASE research posters was displayed throughout the two day event.
Year(s) Of Engagement Activity 2019
 
Description EPSRC Prosperity Partnership Workshop (25/09/18) 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Industry/Business
Results and Impact The industrial PI and Project Manager participated in the EPSRC Prosperity Partnership Workshop, presenting the project to existing and new Prosperity Partnership teams. This generated discussion on best practice in research, both in terms of governance and in the management of collaborative research. New Prosperity Partnership grant holders were informed and able to take advice from the lessons learnt by the projects already underway (including T-B PHASE).
Year(s) Of Engagement Activity 2018
 
Description Hosted EPSRC visit at University of Bristol (14/11/19) 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Other audiences
Results and Impact The T-B PHASE programme was presented to EPSRC representatives when they visited the University in November 2019. This included Amanda Chmura, Deputy Director Business and Impact Partnerships; Liam Boyle, Portfolio Manager in AI and Data Science; Ana Martinez, Portfolio Manager - EPSRC Business Engagement; Steve Webb, Portfolio Manager - EPSRC Business Engagement.
The lead investigators (academic and industry) and Project Manager attended along with the Thales-Bristol partnership managers. Discussions included accrued impact, reporting impact, mid-term review requirements, the EPSRC annual report and the quarterly reports provided, together with a discussion around issues encountered and the sharing of best practice.
Year(s) Of Engagement Activity 2019
 
Description INCOSE UK Architecture Working Group (AWG) (04/02/20) 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Industry/Business
Results and Impact Dr Bugra Alkan (T-B PHASE Postdoctoral Researcher) and Kevin Galvin (Thales Technical Lead - Systems Architecture) participated in the INCOSE UK Architecture Working Group event held at the National Automotive Innovation Centre (NAIC), University of Warwick. The purpose of the activity was threefold: i) advance and evolve the architecture body of knowledge, ii) promote the use and practice of architecture principles in systems engineering and iii) share best practices for the use of architecture in systems engineering projects.
As a new researcher on T-B PHASE, Dr Alkan was invited to participate in order to extend his knowledge in the field of systems engineering, especially in system-of-systems topics, as well as build networks within the INCOSE.
Year(s) Of Engagement Activity 2020
URL https://www.incose.org/incose-member-resources/working-groups/process/architecture
 
Description Interactive Artificial Intelligence (IAI) Symposium (24/09/19) 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Other audiences
Results and Impact Dr Michael Crosscombe (T-B PHASE Postdoctoral Researcher) presented his work on Decision-Making in Robot Swarms to the Interactive Artificial Intelligence (IAI) Symposium at the University of Bristol. Several T-B PHASE team members including the industry PI attended the event, which aimed to showcase new research in interactive AI (e.g. robotics, AI policy) to University academics and their industry partners.
Year(s) Of Engagement Activity 2019
 
Description Thales UK Research Engagement Day (28/11/18) 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Industry/Business
Results and Impact T-B PHASE team members participated in the Research Engagement event at Thales HQ, Reading in November 2018. The event was organised to showcase a diverse range of research and technologies from across Thales UK and its academic engagements, and some 45 topics were exhibited from across the business lines. The event included attendees from Thales, the MoD Chief Scientific Advisor, senior decision makers within the Home Office, a range of senior staff from MoD Main Building and Dstl, senior customers from the rail industry and selected academic partners.

T-B PHASE research was showcased via poster presentations by PDRAs and PhD students from the project team and linked research projects from the University of Bristol.

The Thales-Bristol Strategic Partnership that governs the T-B PHASE programme was highlighted as a key activity for the Thales UK Research, Technology and Innovation division.
Year(s) Of Engagement Activity 2018
 
Description Towards Greater Autonomy in Space workshop (21/01/20) 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Industry/Business
Results and Impact David Harvey (Thales Researcher/Technical Lead on T-B PHASE) attended the University of Liverpool/Space Applications Catapult workshop: Towards Greater Autonomy in Space in London. The event was attended by around 50 people, mostly from industry including some SMEs and comprised short talks and Q&A sessions. The outcome was an increased knowledge of autonomy research within the space sector.
Year(s) Of Engagement Activity 2020
URL https://s3-eu-west-1.amazonaws.com/media.newsa.catapult/wp-content/uploads/2019/11/16095055/TGA-Agen...
 
Description Transport Systems Catapult - Drones: Solutions and Opportunities for the UK Transport Industry (06/03/19) 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach National
Primary Audience Industry/Business
Results and Impact The TSC brought together experts from the drone industry and transport industry professionals to explore the uses and implementation of drone technology in the UK transport industry at a one-day event in Milton Keynes. The event aimed to explore the potential of this emerging industry and to discuss the solutions it can offer for the UK Transport Industry sector. Regulatory issues were also discussed. T-B PHASE PhD student Will Bonnell participated in the event and reported the key messages back to the research team with regards to the potential applications of Hybrid Autonomous Systems (UAVs) and related issues.
Year(s) Of Engagement Activity 2019
URL https://ts.catapult.org.uk/news-events-gallery/events/droneconference/
 
Description University of Bristol Collective Dynamics Seminar (30/10/19) 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Local
Primary Audience Other audiences
Results and Impact Dr Tom Kent (T-B PHASE Postdoctoral Researcher) presented his research to academic staff and students at a University of Bristol Research Group Seminar. The presentation was entitled: "Single Agent Policies for the Multi-Agent Persistent Surveillance Problem". The aim of the seminar was to effectively communicate complex research concepts to an academic audience to inform them of the subject area, and might potentially lead to future research collaboration with colleagues.
Year(s) Of Engagement Activity 2019