Chaotic Exploration and Learning of Locomotion Behaviors

Abstract

We present a general and fully dynamic neural system, which exploits intrinsic chaotic dynamics, for the real-time goal-directed exploration and learning of the possible locomotion patterns of an articulated robot of an arbitrary morphology in an unknown environment. The controller is modeled as a network of neural oscillators that are initially coupled only through physical embodiment, and goal-directed exploration of coordinated motor patterns is achieved by chaotic search using adaptive bifurcation. The phase space of the indirectly coupled neural-body-environment system contains multiple transient or permanent self-organized dynamics, each of which is a candidate for a locomotion behavior. The adaptive bifurcation enables the system orbit to wander through various phase-coordinated states, using its intrinsic chaotic dynamics as a driving force, and stabilizes on to one of the states matching the given goal criteria. In order to improve the sustainability of useful transient patterns, sensory homeostasis has been introduced, which results in an increased diversity of motor outputs, thus achieving multiscale exploration. A rhythmic pattern discovered by this process is memorized and sustained by changing the wiring between initially disconnected oscillators using an adaptive synchronization method. Our results show that the novel neurorobotic system is able to create and learn multiple locomotion behaviors for a wide range of body configurations and physical environments and can readapt in realtime after sustaining damage.

Behavior control in the sensorimotor loop with short-term synaptic dynamics induced by self-regulating neurons

“…For instance, Shim and Husbands (2012) used intrinsic chaos of weakly-coupled central pattern generators to search for a neurocontroller of a quadruped with eight degrees of freedom, and later stored the successful controllers in the connections between the oscillators, using a form of synaptic plasticity. While the same strategy led to a stable forward locomotion of a swimming robot, Shim and Husbands (2012) reported that the behavior of the quadruped broke after some time. However, a stable 18-joints hexapod forward locomotion is achieved using Walknet (Schilling et al, 2013a).…”

“…Examples include tropisms of wheel-driven robots (Hülse and Pasemann, 2002; Smith et al, 2002), biped walking (Manoonpong et al, 2007; Kubisch et al, 2011), active tracking (Negrello and Pasemann, 2008), quadruped locomotion, (Manoonpong et al, 2006; Ijspeert et al, 2007; Shim and Husbands, 2012), hexapod locomotion (Beer and Gallagher, 1992), and swimming robots (Ijspeert et al, 2007; Shim and Husbands, 2012). …”

Neuronal Assembly Dynamics in Supervised and Unsupervised Learning Scenarios

“…EXPERIMENT 2 Plasticity mechanisms are a common feature in the brain and mediate many (if not all) cognitive processes during learning and development (Turrigiano & Nelson, 2004;Masquelier et al, 2009). There is a rich literature exploring models of artificial neuronal networks with some kind of synaptic plasticity in the context of real or simulated agents engaged in a behavioural task (Urzelai & Floreano, 2001;Sporns & Alexander, 2002;Di Paolo, 2003;Edelman, 2007;Shim & Husbands, 2012), but normally the techniques involve the modulation of the electric connections between nodes of the network as a response to the agent's actions and the environment. Here, we explore the way in which neurons and assemblies relate to each other, and how a modulation of this relationship alone, without other plasticity mechanisms, can be exploited to generate adaptive behaviour.…”

Active Shape Discrimination with Compliant Bodies as Reservoir Computers

“…This has resulted in a renewed focus on the form and function of the body. Repeated successes in the exploitation of inherent, often passive, dynamics in automata and robots have demonstrated that much can be gained, in terms of efficiency and simplification of control, when body-brain-environment interactions are balanced and harmonious [24,15,32,36,31]. Pfeifer and Iida [30] introduced the term morphological computation to refer to the way in which a judiciously selected body morphology can be shown to simplify the task of a controller and might therefore be considered to be performing a function analogous to the computational work it renders redundant.…”