Innate Motivation for Robot Swarms by Minimizing Surprise: From Simple Simulations to Real-World Experiments

Abstract

Applications of large-scale mobile multi-robot systems can be beneficial over monolithic robots because of higher potential for robustness and scalability. Developing controllers for multi-robot systems is challenging because the multitude of interactions is hard to anticipate and difficult to model. Automatic design using machine learning or evolutionary robotics seem to be options to avoid that challenge, but bring the challenge of designing reward or fitness functions. Generic reward and fitness functions seem unlikely to exist and task-specific rewards often have undesired side effects. Approaches of so-called innate motivation try to avoid the specific formulation of rewards and work instead with different drivers, such as curiosity. Our approach to innate motivation is to minimize surprise, which we implement by maximizing the accuracy of the swarm robot's sensor predictions using neuroevolution. A unique advantage of the swarm robot case is that swarm members populate the robot's environment and can trigger more active behaviors in a selfreferential loop. We summarize our previous simulation-based results concerning behavioral diversity, robustness, scalability, and engineered self-organization, and put them into context. In several new studies, we analyze the influence of the optimizer's hyperparameters, the scalability of evolved behaviors, and the impact of realistic robot simulations. Finally, we present results using real robots that show how the reality gap can be bridged.

Control of Multiple AUV Systems With Input Saturations Using Distributed Fixed-Time Consensus Fuzzy Control

“…xi − 0.7 sin(υ xi υ yi ), 2.5 cos(t) + 0.1υ 2 xi + 0.5 sin(υ yi ), 2.5 sin(t) + 0.7υ 2 xi + 0.8 sin(υ zi ), 0.5 sin(t) + 0.2υ 3 ϕi , 0.5 cos(t) − 0.2υ 2 θi , 0.5 sin(t) − 0.4υ 3 Φi , ] T , (i ∈ {1, 2, 3, 4}).…”

“…In some applications, the use of single robots or autonomous vehicles has limited efficiency, due to the limitation of sensing, endurance and payload carrying. To increase the efficiency of robots and autonomous vehicles for these applications, a concept of multiple collaborative robotics or swam robotics have been introduced [3]. For underwater environment, multiple collaborative autonomous underwater vehicles (AUVs) have shown their great efficiency for many challenging applications like seabed monitoring, wind turbine inspection, marine debris monitoring and cleaning, etc [4].…”

Distributed Fixed-Time Consensus Control for Multiple AUV Systems with Input Saturations

“…5 sin(t) − 0.5υ 2 xi − 0.7 sin(υ xi υ yi ), 2.5 cos(t) + 0.1υ 2 xi + 0.5 sin(υ yi ), 2.5 sin(t) + 0.7υ 2 xi + 0.8 sin(υ zi ), 0.5 sin(t) + 0.2υ 3 φi , 0.5 cos(t) − 0.2υ 2 θi , 0.5 sin(t) − 0.4υ 3 Φi , ] T , (i ∈ {1, 2, 3, 4}).…”

Reinforcement learning based flight controller capable of controlling a quadcopter with four, three and two working motors

“…In other literature, it used to increase the performance of well-known objectives in robotics. For example, in [12] curiosity is used for motion planning of humanoid robots, in [13] for control of swarm of robots, and in [14] for robot navigation.…”