Rosenkind, Micah, WINSTANLEY, GRAHAM and BLAKE, ANDREW
Creating believable characters using a synthetic psychology approach
In: CGAMES 2008, 3-5 Nov 2008, Light House Media Centre, Wolverhampton, UK, 3-5 November 2008.
The aim of this project is to make simulated characters in interactive real-time software scenarios more believable by increasing the personality of their behaviour. This will be achieved by implementing biologically inspired Artificial Life (Alife) architectures into their control mechanisms. The method to achieve this is inspired by the constructionist “Synthetic Psychology” approach adopted by neurologist Valentino Braitenberg (Braitenberg 1984). The architecture described in this thought experiment is comprised of a network of functional components and is an abstraction of the neural-architecture of human and mammal brains. The objective of this project is to implement a memoryprediction mechanism (Hawkins and Blakeslee 2004) that will enable virtual characters to behave in unpredictable, unique and interesting ways and allow them to become more sensitive to their environment and events, thus further suspending disbelief. In addition to proposing a process for creating simulated characters we also aim to evaluate a method of automatically detecting emergent behaviours exhibited by characters using Kohonen self-organizing maps (SOM). SOMs have proven useful in finding emergent properties in systems from other fields such as usability analysis, marketing and medicine. We believe this capability can assist during the process, which would otherwise require long-term observation for evaluation. Implemented using Microsoft Robotics Developer Studio 2008, this project will produce a series of prototype virtual “Vehicles” based on the designs found in Braitenberg's work, but in some cases amended with more recent findings about the neural architecture of the brain as described by Hawkins (Hawkins and Blakeslee 2004), Bonhoeffer (Engert and Bonhoeffer 1999) and Braitenberg himself.
Actions (login required)
Downloads per month over past year