Towards a Systematic Computational Framework for Modeling Multi-Agent\n Decision-Making at Micro Level for Smart Vehicles in a Smart World

Abstract

We propose a multi-agent based computational framework for modeling\ndecision-making and strategic interaction at micro level for smart vehicles in\na smart world. The concepts of Markov game and best response dynamics are\nheavily leveraged. Our aim is to make the framework conceptually sound and\ncomputationally practical for a range of realistic applications, including\nmicro path planning for autonomous vehicles. To this end, we first convert the\nwould-be stochastic game problem into a closely related deterministic one by\nintroducing risk premium in the utility function for each individual agent. We\nshow how the sub-game perfect Nash equilibrium of the simplified deterministic\ngame can be solved by an algorithm based on best response dynamics. In order to\nbetter model human driving behaviors with bounded rationality, we seek to\nfurther simplify the solution concept by replacing the Nash equilibrium\ncondition with a heuristic and adaptive optimization with finite look-ahead\nanticipation. In addition, the algorithm corresponding to the new solution\nconcept drastically improves the computational efficiency. To demonstrate how\nour approach can be applied to realistic traffic settings, we conduct a\nsimulation experiment: to derive merging and yielding behaviors on a\ndouble-lane highway with an unexpected barrier. Despite assumption differences\ninvolved in the two solution concepts, the derived numerical solutions show\nthat the endogenized driving behaviors are very similar. We also briefly\ncomment on how the proposed framework can be further extended in a number of\ndirections in our forthcoming work, such as behavioral calibration using real\ntraffic video data, computational mechanism design for traffic policy\noptimization, and so on.\n