Combining Vector Symbolic Architecture Aspects and Artificial Intelligence Services Using Edge Deployment
Military / Coalition Issue
Extant battle management systems are inflexible; set up at the start of the mission and unchangeable. Using a dynamic routing environment for the allocations tend to overload the network with overhead. Using machine learning systems on partitioned networks causes them to learn in isolation, when the partition ends, the system will need to discard some of the data, and only use data from one of the partitions (or from the system in the pre-partitioned state).
Core idea and key achievements
We show two methods to make allocations dynamic: centralised software defined coalitions (SDC) control and decentralised vector symbolic architecture (VSA). The centralised SDC uses a top-down approach where users are directed to use a service at a particular location, it uses Machine Learning approaches to ensure that users get the best response times as possible, including using coalition resources when available. If the network is partitioned, the fragments of the network will operate independently, and when the partition ends the system uses a combination of Reinforcement and Transfer Learning to boost recovery times by two orders of magnitude.
Implications for Defence
Using these technologies the coalition can achieve the best of both worlds; centralised control and distributed adaptability.
Readiness & alternative Defence uses
The VSA code is at technology readiness level (TRL)2+ and SDC is at TRL1, we hope during the development of the demo that we can make it more ready.
Resources and references
- Simpkin, Christopher, et al. “A scalable vector symbolic architecture approach for decentralized workflows.” COLLA (2018).
- Simpkin, Chris, et al. “Constructing distributed time-critical applications using cognitive enabled services.” Future Generation Computer Systems 100 (2019): 70-85.
- Simpkin, Christopher, et al. “Dynamic distributed orchestration of Node-Red IoT workflows using a vector symbolic architecture.” 2018 IEEE/ACM Workflows in Support of Large-Scale Science (WORKS). IEEE, 2018.
- Singla, Ankush, Elisa Bertino, and Dinesh Verma. “Preparing network intrusion detection deep learning models with minimal data using adversarial domain adaptation.” Proceedings of the 15th ACM Asia Conference on Computer and Communications Security. 2020.
- Leung, Kin K., et al. “Reinforcement and transfer learning for distributed analytics in fragmented software defined coalitions.” Artificial Intelligence and Machine Learning for Multi-Domain Operations Applications III. Vol. 11746. International Society for Optics and Photonics, 2021.
- Zhang, Ziyao, et al. “Efficient Reinforcement Learning with Implicit Action Space” 4th Annual Fall Meeting of the DAIS ITA, 2020
The underlying vector symbolic architecture code has been published as Open Source.
Imperial College, Purdue University, Cardiff, IBM UK, IBM US, ARL and Dstl