EPSRC-JLR Workshop 9th December 2014
Increasing levels of autonomy of the driving task changing the demands of the environment Increased motivation from non-driving related activities Enhanced interface methodologies enabling smart feedback of threats and hazards Vehicles are driven manually, driver is responsible for the safe control of the vehicle Driver needs to be aware of hazards, managing the demand of the environment Example research challenges: Managing variable levels of driver-vehicle control and defining the handover points considering driver skill, awareness and state What does the information content look like between autonomous and manually driven vehicles? How will the vehicle interior in autonomous vehicles change how users interact in our vehicles? Increasing demands of the wider environment including the connected life of the customer Some basic autonomy for driving (adaptive cruise control, lane keep, queue assistance, emergency stop), interfaces considered as discrete individual elements Human Interaction
Control is achieved by taking sensor data, processing it and actuating brakes, powertrain and steering Features operate as an orphan vehicle covering Safety, Parking and Cruise Predictive capability of other vehicle & infrastructure data Multiple-feature interaction for consistent user-experience Highly automated, discrete functionality Fully autonomous vehicle connected to various modes of transport Example research challenges: Functionality must always be safe in any environment Ability to operate anytime & anywhere (weather, territory & on/off-road) Take account of vehicles with very different levels of autonomy, connection and data sharing capability? How can an autonomous vehicle manage complex junctions, and how does it cope with the fact that traffic protocol varies across the globe? Autonomous Vehicle Control How can the vehicle develop a predictive capability based on user patterns and real-time data? 3 How does the vehicle achieve self-learning, dynamic control based on a holistic, system-level understanding? 2 1
The Self-Learning Car Project The car knows and responds to: who you are, what you want and the environment around you Reduces the amount of driver interaction with vehicle required Applies to areas such as: climate, audio, phone Autonomous Intelligent Features Comfort and convenience features: Auto lights Auto wipers Advanced Driver Assistance (ADAS) features: Automatic cruise control Auto parallel park Automatic emergency braking Example research challenges: Driver and passenger identification Developing machine learning algorithms to recognize driver routine Developing algorithms that are robust to changes and exceptions to customer routine Making inferences based on new sources of information and sensors Reliability of connectivity restricts features TO WARDS AUTONOMY S M A RT A N D C O N N E C T E D C O N T R O L
Product Development / Manufacturing data Vehicle system data Driver Inputs Diagnostic data Vehicle failure prediction Data fusion and analytics Optimisation of vehicle component design Optimisation of user interface Location, environmental conditions Access to quicker and richer data sources via vehicle telematics connections gives lots of opportunities Quicker identification and prediction of vehicle faults Optimisation of vehicle component and system design to match real-life customer usage and environmental condition (opportunities for weight/costs/c02 reduction) Until recently, virtually all data on the vehicle is not stored or transmitted (limited data is downloaded during a vehicle service) Data is usually therefore difficult to access and sparse Conventional analytic techniques are used to analyse data (e.g. fault causes) Data Fusion & Analytics However, vehicles are increasingly connected to the cloud, which allows transfer of information (e.g. diagnostic data) without waiting for a service Example research challenges: Developing algorithms/techniques to predict vehicle faults from data on vehicle usage, technical data and environmental conditions Prediction models that adapt to local market usage, updates to the vehicle, differences between vehicles Aggregating data from multiple vehicles/markets/driver types in order to find patterns across these parameters Communicating and presenting the results to both technical and non-technical users TO WARDS AUTONOMY S M A RT A N D C O N N E C T E D C O N T R O L
Radars front, rear and corner Cameras front, rear & side Ultrasonics front & rear More sensors on the vehicle including FIR, V2X & ehorizon Managing conflicts in sensed information Second generation sensing - sensors with built-in intelligence Sensor interoperability modelling Example research challenges: Prediction and anticipation of unknown events e.g. seeing round bends in the road Robust environmental data fusion & which sensors to believe when they are giving conflicting information? All-weather and lighting capability Longer range sensors & connection to V2X Optimisation reduced number of sensors Robust driver state and health Sensing
Add image Where we are now: Vehicles are cloud connected Reliability of connectivity restricts features Drivers and passengers want to access the connected services they use Future Directions: - 5G impact on vehicle and connected landscape - How will the vehicle integrate with Internet of Things - Requirements on the Infotainment in an autonomous vehicle Off-Board Data Sharing & Infrastructure Example research challenges: What vehicle functions can be cloud based? What is the fall back if connectivity is lost? How will the vehicle be integrated with smart cities? How will the user s digital ecosystem be integrated into the vehicle experience? How will the digital media experience in the vehicle be supported?
Vehicles are complex distributed computer systems - managing compellability is a major issue Limited connectivity in and out of vehicle. A ridged closely coupled architecture validated as a system Distributed Systems & Control Architecture Example research challenges: What vehicle functions can be cloud based? What is the fall back if connectivity is lost? How do we run software components with different integrity levels on the same hardware? How do we manage the additional threats created by greater connectivity? What role will consumer electronics play in the vehicle of the future? Functional growth Utilising printed electronics Architecture consolidation, a smaller number of powerful engine control units Migration towards a service orientated architecture Connectivity extends beyond infotainment domain TO WARDS AUTONOMY S M A RT A N D C O N N E C T E D C O N T R O L
Prototype vehicle based testing for embedded software Paper based robustness system assessment Expensive vehicle tests to meet safety regulations Reduce dependency to physical prototypes Methods, tools and techniques for offline software development Data analysis and system modelling Example research challenges: Modelling of vehicle external environment including connectivity Data analysis and trust (human, system) Abstract modelling of digital eco-systems Verification, Robustness & Safety TO WARDS AUTONOMY S M A RT A N D C O N N E C T E D C O N T R O L
Progression to higher levels of autonomy Current production systems are partially automated There will be a lower level of supervision from the driver There will be limitations on autonomy unless the law is changed Example research challenges: Legislative changes & accommodation of global legislative differences Cruise & Parking features require driver supervision Product liability - how does the industry protect itself? E.g. U.S. punitive damages Legislation is driving automated safety features - (Discovery Sport achieved EuroNCAP 5-Star rating with AEB) International or national standards (ISO, CEN, SAE etc.) to be established Justification of ethics employed Data protection when intelligent features & analytics are introduced Legislative Framework Add image Worldwide 1.24 M deaths PA TO WARDS AUTONOMY S M A RT A N D C O N N E C T E D C O N T R O L
EPSRC-JLR Workshop 9th December 2014