Distributed and decentralized mixed-integer programming for collaborative autonomous driving

• Grant number: 24/04703-0
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: August 01, 2024
• End date: July 31, 2026
• Field of knowledge: Engineering - Electrical Engineering - Industrial Electronics, Electronic Systems and Controls
• Principal Investigator: Janito Vaqueiro Ferreira
• Grantee: Angelo Caregnato Neto
• Host Institution: Faculdade de Engenharia Mecânica (FEM). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

This research project addresses the problem of designing distributed or decentralizedplanning modules for collaborative autonomous vehicles using Mixed-Integer Programming(MIP). In this class of mathematical programs, both continuous and integerdecision variables are optimized, allowing the solution of several multi-agent problemsranging from motion planning and connectivity maintenance to task allocation. Mixedintegerprogramming has also been integrated into receding horizon strategies throughModel Predictive Control with Mixed Integer Programming encoding (MPC-MIP),yielding a powerful technique to coordinate multiple agents in dynamic environments.In spite of the recent success of MIP-based multi-agent motion planners, the majorityof proposals employ centralized architectures that impose limits on the scalability andcomplexity of the addressed problems. Consequently, demonstrations have been limitedto small-scale or simplied engineering scenarios. This research project addresses thisgap between theory and practice by investigating the viability of distributed and decentralizedmixed-integer methods in the design of motion planning and decision-makingmodules considering the concrete engineering challenge of collaborative autonomousdriving. The proposed case study is the collaboration of multiple autonomous agents,modeled as the autonomous vehicle platform from the State University of Campinas,in realistic trac conditions. Through extensive simulation evaluations and, if possible,experimental demonstrations using small-scale vehicular platforms, this projectaims to demonstrate the potential of distributed or decentralized MIP and MPC-MIPplanners as reliable and scalable techniques for multi-vehicle autonomous coordination.