Details of Research Outputs

It is expected that autonomous cars will be the major form of transportation in the next few decades. However, the major obstacle to the success of autonomous cars is the safety issue which can be classified as internal and external. In this study, we focus on internal safety issues such as hardware and software issues. Due to the limitations of the hardware and software, it is very difficult for autonomous cars to detect the full range of system potential faults by themselves. Each car has weaknesses. The cars may encounter different defects such as outdated software, malfunction of sensors, etc. We found that we can take advantage of Condorcet's jury theorem to solve the problem: 'juries consisting of many individuals are likely to reach better decisions than single experts'. To implement the idea, we proposed a self-reflection algorithm by using P2P benchmarking, so that autonomous cars can proactively diagnose their system performance regularly, and detect all potential faults by using the collective views of peers. The proposed algorithm also brings other advantages such as resisting the occasional errors, adapting to the evolution of technological changes, easily extending the processing capacity, etc. The simulations showed that the results are promising.