Vehicles often rely on a set of sources to determine location; these are typically GPS or inferential reference solutions. As part of several studies the centre is exploring the option of using other data sources on vehicles to determine position based on other environmental cues. One option being explored is imagery, using this to determine object layout, building position and so on to further refine or locate the vehicle. Solutions such as this have been used in rail systems to improve vehicle location accuracy and have been proven to be highly effective.
The centre is exploring different types of vehicle communication and the impact loss of communications paths or loss of integrity in those data affects overall resilience of Connected Autonomous Vehicles (CAV).
|
Real World vs Training World for AIThe centre is exploring the resiliency of AI algorithms. One approach being explored is the concept of numerically quantifying the difference between the real world and the simulation or “world” the AI was trained using. It is not uncommon for AI systems that are extensively trained on large data sets (particularly those that are machine learning based) to exhibit strange behaviour when faced with a real world non-standard situation. The approach being explored is a means to compare what the AI is ‘experiencing’ with the real world information and make a comparative assessment to determine “is this something similar to what I’ve seen before or not?” and then use this comparison to influence AI decision making. |
This is one of the core concepts used by ResilientWorks, and is the start of a new innovative approach to engineering generally. As systems become more complex, so too does the ability to effectively categorise or precisely define their behaviour for acceptance or regulatory purposes, as the possibility of emergent properties simply makes it too difficult to be sure that testing a given system state gives an accurate impression of its total performance or behaviour. This is of course how many systems are designed and accepted, requirements are created, results expected and tested for and this forms the basis of evidence a system is safe, reliable and fit for purpose. What Cyber Resilience considers, is the concept of accepting behaviour and not specific system outputs. Is the behaviour right? Is the solution behaving as we expect in a set of given circumstances? Is that behaviour safe? Rather than imposing exact limits and constraints to system design and outputs this means platforms can be different, can behave slightly differently, but ultimately behave in a somewhat expected manner. ResilientWorks looks at how this concept can be used to re-examine how we look at system design proving (how we evidence that a system behaves as intended or behaved as it should in the past) can be evolve into something more usable, focusing on design resilience rather than absolutes.
As part of a larger modelling of electric vehicles becoming part of a micro grid ResilientWorks is examining battery management options and the data that might be required to support energy aggregation systems in the future. This might be an expansion to current battery management systems to allow them to communicate with energy aggregators upstream as part of a collective pool.
Vehicle to Grid ChargingThis project examines how the CAV/EVs could interact with the distribution networks in a way that supports balancing supply and demand but also offers additional benefits, not only for the energy providers but also for the vehicle owners. The vehicle has a need to manage its batteries, understand its consumption and provision capacity and balance this with the current health status of the battery. The electricity network needs to be balanced and could use vehicle batteries to do this. |
|
The availability and reliability of electric vehicle charging points will become more important as electric vehicle uptake increases. Charging at home, work and leisure locations along with ‘in-transit’ options need to be implemented in greater capacity.
Some locations will be limited by network capacity and will influence the number of charging bays and the rate at which charging will take place without a local electricity network upgrade. Identifying options for ‘buffer storage’ to manage peaks in demand will become part of the system optimisation.
Blending vehicle charging with other services such as dinner and a movie offers the chance to forecast charging demand more reliably and make charging more commercially attractive to the consumer.
Micro-Grids afford an opportunity to take vehicle charging away from a direct demand on the electricity network. Allowing for energy management across a range of consumers such as offices, warehousing and factories in addition to vehicle charging. This allows for greater flexibility in the options to ‘balance’ the system through high and low demand periods by integrating managed energy storage.
This storage could come from a combination of dedicated storage batteries and vehicle batteries connected to the micro-grid.
An aggregation of multiple micro-grids could also be leveraged to supply electricity back into the electricity network in order to manage network availability, reliability and quality of supply.