Toshiba Deploys Quantum-Inspired Optimization on Robot
Toshiba and MIRISE have deployed a quantum-inspired optimization computer on an autonomous mobile robot. The achievement marks a step toward using such optimization techniques for real-time decision-making in embedded systems, foreshadowing potential applications in latency-sensitive financial trading environments.
- The core technology, Toshiba's Simulated Bifurcation Machine (SBM), is a quantum-inspired optimization computer designed to solve complex combinatorial optimization problems at high speed. Unlike true quantum computers, it runs on classical hardware like FPGAs, making it suitable for embedded, real-time applications with power and cost constraints. - This is not Toshiba's first application of SBM in finance; in December 2023, the company announced it had successfully used the SBM to develop and execute high-speed stock trading strategies on the Tokyo Stock Exchange. That system was designed to analyze complex correlations between the prices of up to 2,000 Japanese stocks to identify trading opportunities. - The use of an FPGA (Field-Programmable Gate Array) is critical for the low-latency aspect mentioned. In high-frequency trading (HFT), FPGAs are used to accelerate systems by processing market data feeds and executing trading logic directly in hardware, significantly reducing latency compared to software-based approaches. This can result in order execution times in nanoseconds. - The collaborator, MIRISE Technologies, is a joint venture established in 2020 between automotive parts manufacturer DENSO (51% ownership) and Toyota Motor Corporation (49% ownership). Its primary mission is the research and development of next-generation in-vehicle semiconductors for electric and autonomous vehicles, focusing on power electronics, sensing, and System-on-a-Chip (SoC) solutions. - The underlying challenge being solved is a combinatorial optimization problem, which is common in finance for tasks like portfolio optimization, risk management, and algorithmic trading. These problems involve finding the best solution from a vast number of possible combinations, which is computationally intensive for classical computers. - Toshiba has been actively developing its SBM technology for financial applications, including a joint experiment with Dharma Capital in 2021 to apply it to high-frequency trading strategies for Japanese stocks. The goal was to uncover previously undiscovered arbitrage opportunities by solving combinatorial optimization problems faster than conventional methods. - The SBM technology has been made accessible to developers through cloud platforms like Azure Quantum, where it is offered as an Ising model solver capable of handling problems with up to 100,000 variables. This allows for experimentation and application in areas like dynamic portfolio and risk management without needing specialized hardware.
