The aim of the project is to develop an intelligent system for maneuver of on-road autonomous vehicles. We propose a hybrid intelligent framework using tools from statistical learning theory and predictive control techniques.
With a viewpoint of reducing accidents owing to human errors, boosting passenger safety, optimizing travel time and fuel consumption, increasing traffic throughput and enhance road space utilization, there is considerable interest in developing autonomous vehicles in the last decade. While a section of researchers have investigated the problem using learning strategies leading to Artificial Intelligence (AI), another section of researchers have used contrasting ideas from classical control theory.
Knowledge of dynamics of different systems and sub-systems of a vehicle allows developing autonomous controllers in the strata of control theory. But performance fluctuations and deviation from the predicted outcome owing to continuous parameter variation limits the utility of these controllers to be applied for real-time applications. In any mechanical system, parameter variations arise mainly due to wear of components. Therefore, robust controllers ensuring stability of the system are difficult to develop. Moreover, the domain of operation is highly constrained. AI can be deployed as an alternative to autonomous control of vehicles and has shown quite satisfying results for different traffic scenarios. But once again, performance of AI is compromised owing to the extremely stochastic nature of the environmental as well as surrounding traffic characteristics. For cases like wet-road, icy road or split-μ conditions AI fails to ensure stability of the vehicle as it is impossible to learn all possible outcome in these conditions and passenger safety is in jeopardy. In these scenarios, the controller based on the first-principle models of vehicle can play an important role in conjunction with AI.
Through this project we propose to integrate the learning based AI and model-based predictive control strategies to improve performance of autonomous vehicles.