In the realm of autonomous systems, efficiency and adaptability are key factors in determining success. This case study explores two projects that utilized Evolutionary AI, specifically Liquid Time-Constant Networks (LTCs), to enhance the autonomy of an airborne quadrotor drone and a land-based vehicle. The results demonstrate the superiority of Evolutionary AI over Conventional AI in terms of energy efficiency, resource utilization, and adaptability
Project 1
Quadrotor Drone
Flight Time Extension
Increase the flight time of a quadrotor drone
Substituted CPU-based processing with the Vivum IP core on an FPGA board
Employed Evolutionary AI (LTC – CTRNN)
Extended flight time by 11.66%
Project 2
Land-Based Vehicle
Range Comparison
Nvidia Pegasus
Self-driving chip (S1) running on Conventional AI for sensor fusion
Vivum IP core
66.65% additional battery/miles (from 232 miles to 387 miles)
40.04% additional battery/miles (from 232 miles to 324 miles)
Advantages
Evolutionary AI in Autonomous Systems
These case studies showcase the potential of Evolutionary AI in enhancing the autonomy of various systems. By demonstrating superior energy efficiency, resource utilization, and adaptability compared to conventional AI, Evolutionary AI proves its effectiveness in real-world applications. The projects successfully met the requirements set by the sponsors, highlighting the potential of Evolutionary AI in pushing the boundaries of autonomous systems.
As the demand for efficient and adaptable autonomous systems continues to grow, Evolutionary AI presents a promising approach to tackle the challenges faced by industries such as drone technology and autonomous vehicles. The ability to optimize performance while minimizing resource consumption positions Evolutionary AI as a key player in shaping the future of autonomous systems.