Project No. 2305
Dr Jimena Berni – University of Sussex
Dr Carlo Tiseo – University of Sussex
Dr Melissa Andrews – University of Southampton
Biological systems have evolved for aeons in symbiosis with the environment, and successful strategies have been kept during evolutions across multiple species.
For example, the circuits controlling rhythmic movement (CPG) are present in the spinal cords of insects and humans and can generate efficient search behaviours.
Earlier studies on fruit fly larvae show how the patterns of neuronal activity are modulated to promote foraging exploration in unknown environments. In contrast, our artificial systems (e.g., robots) can navigate discretely well in known conditions due to efficient computational algorithms developed in the last few decades; however, they highly rely on environmental modelling, which is not available for unknown environments. Furthermore, these methods are not energy efficient.
This project explores new methods to design a robot capable of solving a foraging task in an energy efficient way. We will implement patterned stereotyped strategies in an artificial system inspired by our experimental observation of fruitly larvae while severely limiting the computational power available on the system to minimise energy consumption. We will use two methods: a traditional microcontroller technology, and it will explore the algorithmic implementation of a network of coordinated nonlinear oscillators mimicking the CPG. The second platform will use electronics based on memristor, capable of reproducing the neuronal activity to generate a hardware implementation of a CPG equivalent of our system. The goal of our multidisciplinary team is to reduce at least 10 times the computational cost of solving this type of problem.