“Nothing is too wonderful to be true if it be consistent with the laws of nature; and in such things as these, experiment is the best test of such consistency.”
– Michael Faraday
We explore Bio-inspired algorithms/mechanisms for fusing Soft (Mobile and Static) Agents with their hard Counterparts viz. Robots and Devices to form an Intelligent IoT/CPS.
We envisage this Lab. to be home to enthusiastic practical researchers in real-world bio-inspired robotics. We believe only in actual implementations and try to ward ourselves from just mere models, simulations and closed world scenarios – flaws that led to the decline of the classical (theoretical) AI. We use Non-Standard Computing (NSC) paradigms to evolve the world of real robots and devices. This lab. was partially funded by the FIST programme of the Department of Science & Technology, Government of India.
The main research focus is to build an Artificial Being (ABe) which unlike humanoid robots, is a congregation of networked mobile robots emulating the various parts/organs of a body. The ABe, which is a Cyber Physical System, is still in the making and when done will manifest in the form of a small group of wireless mobile robots, sensor nodes and connected devices that will maintain their integrity, connections, spread out, re-group, share intelligence autonomously and even interface with human beings and comprehend their emotions. Information and hence intelligence is shared using autonomous mobile agents that knit through these robots, serving them based on demand. These heterogeneous agents function in a distributed manner, cloning as and when required to hasten information flow while at the same time controlling their population within the network using stigmergy thus saving on precious bandwidth.
Mechanisms for agent mobility amongst networked robots and devices that request information using pheromone diffusion (not to be confused with the theoretical ACO model) and stigmergy based clonal population control have been tested and implemented. Artificial Immune Systems (AIS) based learning has also been implemented. Learning using on the fly sharing, thanks to the use of mobile agents has also been demonstrated. Behaviour arbitration using emotion recognition and generation has also been attempted. Emotion based memories have also been realized to augment the control and behaviour of situated robots.
On-The-Fly Programming (OTFP) using mobile agents improvised using Tartarus, an agent based platform developed at this Lab., aid in realizing Cyber Physical Systems which are both flexible and scalable.
We also investigate non-traditional approaches to language processing – both text and speech. Of late we have been able to use Jerne’s Nobel prize winning Idiotypic Network to formulate an algorithm that can generate correct sentences. The work is based on the comments made by Jerne in his Nobel lecture on the Generative Grammar of the immune system.
In all our approaches we strongly recommend techniques found in Nature (Bio-inspirations) since we feel that they are the most rugged and time-tested ones and thus are the best for use in real world applications. This forces us away from traditional incremental research and explore new avenues by treading on untrodden paths.
The research has yielded other valuable side-products which include –
- Intelligent Cyber Physical Systems: The ABe itself is a Cyber Physical System (real network) comprising mainly mobile agents, robots for actuation and devices (processing units, sensors, etc.). Intelligence is distributed, decentralized, ubiquitous and assimilated, shared and disseminated across the network by the mobile agents.
- Practical Implementations of both Clonal Selection and a Real Distributed Decentralized Localized Idiotypic Network (LIN) (using Typhon): We portray and emulate how the actual distributed, decentralized biological immune system possibly functions within our body using Clonal Selection and Jerne’s Idiotypic model. A set of mobile agents (antibodies) form the Idiotypic Network which services a set of nodes (which are attacked by antigens) comprising a real network.
- Synthetic Multi-Agent Emotion Generator: While many have worked on reflecting emotions on robotic faces, we have realized an emotion generator that accepts percepts from the real environment and generated synthetic emotions much the same way as humans do, which eventually can be used to control the actions and behaviours of a robot. The generator features adrenaline-like responses, emotional memories, emotion charging, decay and persistence.
- Typhon†: A Mobile Agent Emulation Platform (MAP) running on LPA Prolog’s Chimera Agent System which supports several features for mobility, dynamic on-the-fly code alteration, network management, partial security, etc. The same can be used as a tool to emulate and evolve, among others, real distributed and decentralized bio-inspired algorithms. Know more by clicking here and download it here.
- Tartarus: A new and faster version of Typhon based on SWI-Prolog with new features including threaded agents, is now available for download and use. It can be used in conjunction with embedded systems such as the Raspberry Pi, Intel’s Galileo boards and also integrate robots to form a Cyber-Physical System. Currently we have used it to emulate better ways for flow of emergency vehicles and also to integrate robots with Cyber-Physical Systems. (Click here to check for published work using Tartarus.)
- PherCon-C: This is a bidirectional distributed search mechanism between an agent and a real networked node that requires the former’s service. It couples conscientious search with pheromone diffusion and cloning, to facilitate fast, efficient mobile agent migration within a Cyber physical Systems/Internet of Things (IoT). The mechanism is robust and can perform well even in dynamic networks where the robotic nodes/devices are mobile.
- Lego NXT® Interface: LPA Prolog interface for Lego NXT® robots which allows for real world control of robots using Prolog/Typhon. This allows for testing algorithms with real robots in the loop. The interface supports control of multiple and remotely deployed robots over a network. A similar interface for Tartarus has also been developed.
Key Faculty:
†Not to be confused with typhoon! In Greek mythology, Typhon is known to be the father of all monsters, including Chimera.