Biologically-inspired control framework for insect animation

• Main Author: Guo, Shishi
• Published: Bournemouth University 2015
• Subjects: 302.2
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.667835

Insects are common in our world, such as ants, spiders, cockroaches etc. Virtual representations of them have wide applications in Virtual Reality (VR), video games and films. Compared with the large volume of works in biped animation, the problem of insect animation was less explored. Their small body parts, complex structures and high-speed movements challenge the standard techniques of motion synthesis. This thesis addressed the aforementioned challenge by presenting a framework to efficiently automate the modelling and authoring of insect locomotion. This framework is inspired by two key observations of real insects: fixed gait pattern and distributed neural system. At the top level, a Triangle Placement Engine (TPE) is modelled based on the double-tripod gait pattern of insects, and determines the location and orientation of insect foot contacts, given various user inputs. At the low level, a Central Pattern Generator (CPG) controller actuates individual joints by mimicking the distributed neural system of insects. A Controller Look-Up Table (CLUT) translates the high-level commands from the TPE into the low-level control parameters of the CPG. In addition, a novel strategy is introduced to determine when legs start to swing. During high-speed movements, the swing mode is triggered when the Centre of Mass (COM) steps outside the Supporting Triangle. However, this simplified mechanism is not sufficient to produce the gait variations when insects are moving at slow speed. The proposed strategy handles the case of slow speed by considering four independent factors, including the relative distance to the extreme poses, the stance period, the relative distance to the neighbouring legs, the load information etc. This strategy is able to avoid the issues of collisions between legs or over stretching of leg joints, which are produced by conventional methods. The framework developed in this thesis allows sufficient control and seamlessly fits into the existing pipeline of animation production. With this framework, animators can model the motion of a single insect in an intuitive way by specifying the walking path, terrains, speed etc. The success of this framework proves that the introduction of biological components could synthesise the insect animation in a naturalness and interactive fashion.