Neuroevolution of Inverted Pendulum Control: A Comparative Study of Simulation Techniques

Evolutionary Robotics Controllers with Proprioception Facilitated by Neural-Network-Based Simulators

“…These NNs were termed simulator neural networks (SNNs). Such simulators have been used and demonstrated to be effective [39,40,47]. This process of simulator creation requires less knowledge of a robotʼs dynamics than a mathematics-based approach.…”

Complex Morphology Neural Network Simulation in Evolutionary Robotics

“…The use of SNNs for simulating the behaviours of an inverted pendulum robot has been successfully demonstrated. 4,16 SNNs can serve as a fast surrogate model to a more high-fidelity simulation where evaluations can be time-consuming. 19 Prior investigations into the use of SNNs for simulating the behaviours of a complex snake-like robot required structured changes to the joint angles.…”

“…For certain problems, SNNs may be faster, more accurate and easier to construct than physics-based approaches. 16,23 This is due to SNNs being relatively simple and behaviourally focused rather than built from specialised assumptions. SNNs purposely avoid many of the mathematical complexities required in ER when designing simulators.…”

“…In comparison, model-fitting approaches based on empirical data require little specialised knowledge and are relatively simple to develop. 16,17 Physic-based simulators model the various underlying features between the robot and the environment. Alternatively, empirical models can be constructed and optimised from empirically collected data.…”

“…SNNs are known to enjoy good generalisation abilities, computational efficiency and noise tolerances; require little specialised knowledge to construct; and can result in good transferability of robot behaviours from simulation to reality. 5,10,16 Prior work investigating the use of SNNs as an alternative to physics-based simulators in ER focused on simple, low-dimensional problems. No known prior research has investigated the effectiveness of SNNs on complex, high-dimensional problems in ER.…”