FPGA implementation of mixed integer quadratic programming solver for mobile robot control

Abstract

We propose a high-speed mixed integer quadratic programming (MIQP) solver on an FPGA. The MIQP solver can be applied to various optimizing applications including real-time robot control. In order to rapidly solve the MIQP problem, we implement reusing a first solution (first point), pipeline architecture, and multi-core architecture on the single FPGA. By making use of them, we confirmed that 79.5% of the cycle times are reduced, compared with straightforward sequential processing. The operating frequency is 67 MHz, although a core 2 duo PC requires 3.16 GHz in processing the same size problem. The power consumption of the MIQP solver is 4.2 W.