Incorporation of Swarm Intelligence in Autonomous Cars

Currently, swarm robotics is one of the most promising technologies in computer science. As technologies become smarter, the potential for swarm robotics improve and the coordination of multi robot system finds applications in various fields. This paper adds new perspective to autonomous cars by proposing to incorporate swarm robotics into it. This paper summarizes various algorithms used in swarm robotics that can also be used in autonomous cars. The current working of autonomous cars would tend to be inefficient in future when streets would be populated with them, by using our idea the efficiency would be improved massively. Keywords—Autonomous Cars, Swarm Robotics, Swarm Intelligence, Swarm Algorithms, Google cars, Autonomous cars. INTRODUCTION Swarm robotics finds its applications in various fields ranging from space exploration to bomb defusal. The ability to divide tasks and find a collective solution to a particular problem is the highlight of swarm robotics. Various algorithms have been written to improve the collective behavior of a swarm. Autonomous cars have retained the limelight, thanks to Google cars. The success of autonomous cars would result in a large number of selfdriven cars on our streets in the near future. The concept of swarm robotics and autonomous cars are closely related and combining them would only result in improved efficiency in the future. I. SWARM INTELLIGENCE Swarm Intelligence systems are typically made up of a population of simple agents interacting locally with one another and with their environment. The group of individuals acting in such a manner is referred to as a swarm [1]. The main objective of swarm intelligence is to aggregate the individual behavior, interactions with the neighboring robots and the interactions with the environment to achieve a collective behavior that can be used to solve problems collectively. For the swarm to be more intelligent, the swarm should work more congruently. A. Introduction to Swarm Intelligence Algorithms Most well-known algorithms to implement Swarm Intelligence are Ant Colony Optimization and Particle Swarm Optimization. 1) Ant Colony Optimization: The ant colony optimization as the name suggests is based on the collective behavior of the ant colony. Considering the collective work of an ant colony and not an individual ant suggests how problems of survival can be solved. In order to demonstrate this concept Deneubourg conducted the Binary Bridge Experiment [2]. In this experiment ants take two different paths to one particular food source, the ant that returns back first to the colony is the one taking the shorter path. Pheromone (ants deposit pheromone while walking) concentration on the shorter path will be higher since more ants would complete the loop through the shorter path as compared to the longer path. This further encourages the ants to take the shorter path as a result of their social behavior (of following pheromone trail). As a result all ants start to take the shorter route, thus, social interaction and coordination for foraging occurs indirectly through pheromone deposits which modify the environment. Similar concept is used in Swarm robotic algorithm using ACO, a virtual pheromone is deployed to improve the problem solving efficiency, as the task matures more and more efficient solutions may be observed. 2) Particle Swarm Optimization [3]: Particle Swarm Optimization is mostly associated with the bird flocking analogy. Imagine a flock of birds circling over an area where they can smell a source of food. The bird that is closest to the food source chirps louder and all other birds fly around his direction, as soon as another bird comes even more closer to the target, it will chirp louder that the first bird and hence will result in all the other birds flying in his direction. This pattern continues until one of the birds reaches the target (food). Using this analogy PSO is explained, over a number of iterations, a group of variables (birds) have their values adjusted, such that with each adjustment a more efficient solution (target food) is obtained. The adjusted value would be closer to the member whose value is closest to the target. II. EXISTING ALGORITHMS IN SWARM ROBOTICS: A. Shape formation Algorithm in a swarm: Situations might arise where individual agents in a swarm might want to align themselves in a line, or aggregate themselves into shapes, of varying forms and sizes. The SHAPEBUGS Algorithm employs a decentralized approach to achieve swarm formations using local interactions. The algorithm is flexible in that it continues to work even in the face of an unprecedented influx or exodus of agents. The first process of the SHAPEBUGS Algorithm uses trilateration coupled with a gradient algorithm to help the agent locate itself in the arena. A successful instance of execution of the Gradient Algorithm depends on the initial starting position, an Yadhu Prakash et al, / (IJCSIT) International Journal of Computer Science and Information Technologies, Vol. 5 (5) , 2014, 6307-6309