Arup Kumar Sadhu

Multi-Robot box-pushing using non-dominated sorting bee colony optimization algorithm

The paper provides a new approach to multi-robot box pushing using a proposed Non-dominated Sorting Bee Colony (NSBC) optimization algorithm. The proposed scheme determines time-, energy- and friction-optimal solution to the box-pushing problem. The performance of the developed NSBC algorithm is compared to NSGA-II in connection with the box-pushing problem and the experimental results reveal that the NSBC outperforms NSGA-II in all the experiments.

A modified Imperialist Competitive Algorithm for multi-robot stick-carrying application

The paper proposes a novel evolutionary optimization approach of solving a multi-robot stick-carrying problem. The problem refers to determine the time-optimal trajectory of a stick, being carried by two robots, from a given starting position to a predefined goal position amidst static obstacles in a robot world-map. The problem has been solved using a new hybrid evolutionary algorithm. Hybridization, in the context of evolutionary optimization framework, refers to developing new algorithms by synergistically combining the composite benefits of global exploration and local exploitation capabilities of different ancestor algorithms. The paper proposes a novel approach to embed the motion dynamics of fireflies of the Firefly Algorithm (FA) into a socio-political evolution-based meta-heuristic search algorithm, known as the Imperialist Competitive Algorithm (ICA). The proposed algorithm also uses a modified random-walk strategy based on the position of the candidate solutions in the search space to effectually balance the trade-off between exploration and exploitation. Thirteen other state-of-art techniques have been used here to study the relative performance of the proposed Imperialist Competitive Firefly Algorithm (ICFA) with respect to run-time and accuracy (offset in objective function from the theoretical optimum after termination of the algorithm). Computer simulations undertaken on a well-known set of 25 benchmark functions reveal that the incorporation of the proposed strategies into the traditional ICA makes it more efficient in both run-time and accuracy. The performance of the proposed algorithm has then finally been studied on the real-time multi-robot stick-carrying problem. Experimental results obtained for both simulation and real frameworks indicate that the proposed algorithm based stick-carrying scheme outperforms other state-of-art techniques with respect to two standard metrics defined in the literature. The application justifies the importance of the proposed hybridization and parameter adaptation strategies in practical systems. Multi-robot stick-carrying problem is solved by the proposed ICFA.ICFA is fusion of motion dynamics of Firefly and Imperialist Competitive Algorithm.Modified random-walk strategy is proposed to balance exploration/exploitation.Simulation results confirm efficiency of the proposed ICFA in the state-of-art.Experiment with twin Khepera-II mobile robots is done amidst static obstacles.

Arduino based multi-robot stick carrying by Artificial Bee Colony optimization algorithm

Cooperation of the multi-robots is an upcoming appealing area of research in the field of robotics. In this paper, two arduino based mobile robots are carrying a stick by cooperation towards their goal avoiding obstacles. The path planning algorithm is designed with the help of Artificial Bee Colony Optimization (ABCO) algorithm which chooses the optimized path by minimizing the distance between the robots and maximizing the distance from the obstacles. The ultrasonic sensors, encoder, 3-axis compass and XBee module are embedded in the robot to detect obstacle in the path of the robots, the distance travelled by the robot, calculate the direction (coordinate) of the robot and to communicate with other robots respectively. We have also designed our algorithm with the help of differential evolutionary (DE) algorithm. Analyzing the performance of ABCO and DE algorithms, it is observed that ABCO outperforms DE in real-robot experiment with respect to distance metric.

Multi-robot Box-Pushing Using Differential Evolution Algorithm for Multiobjective Optimization

The paper provides a new approach to multi-robot box pushing using a proposed Differential evolution for multiobjective optimization (DEMO) algorithm. The proposed scheme determines time-, energy- and friction sensitive-optimal solution to the box-pushing problem. The performance of the developed DEMO algorithm is compared to NSGA-II in connection with the given problem and the experimental results reveal that the DEMO outperforms NSGA-II in all the experimental set-ups.

Improving the speed of convergence of multi-agent Q-learning for cooperative task-planning by a robot-team

Abstract Learning-based planning algorithms are currently gaining popularity for their increasing applications in real-time planning and cooperation of robots. The paper aims at extending traditional multi-agent Q-learning algorithms to improve their speed of convergence by incorporating two interesting properties, concerning (i) exploration of the team-goal and (ii) selection of joint action at a given joint state. The exploration of team-goal is realized by allowing the agents, capable of reaching their goals, to wait at their individual goal states, until remaining agents explore their individual goals synchronously or asynchronously. To avoid unwanted never-ending wait-loops, an upper bound to wait-interval, obtained empirically for the waiting team members, is introduced. Selection of joint action, which is a crucial problem in traditional multi-agent Q-learning, is performed here by taking the intersection of individual preferred joint actions of all the agents. In case the resulting intersection is a null set, the individual actions are selected randomly or otherwise following classical multi-agent Q-learning. It is shown both theoretically and experimentally that the extended algorithms outperform its traditional counterpart with respect to speed of convergence. To ensure selection of right joint action at each step of planning, we offer high rewards to exploration of the team-goal and zero rewards to exploration of individual goals during the learning phase. The introduction of the above strategy results in an enriched joint Q-table, the consultation of which during the multi-agent planning yields significant improvement in the performance of cooperative planning of robots. Hardwired realization of the proposed learning based planning algorithm, designed for object-transportation application, confirms the relative merits of the proposed technique over contestant algorithms.

Person identification using Kinect sensor

A simple and easy-to-use system is designed for recognition of persons using their walking style. Here Kinect sensor is employed to generate the twenty body joint co-ordinates of the persons. In this proposed algorithm, we have processed five walking data sets from each twenty five persons. Total nine body joint co-ordinates are considered for each frame using Kinect sensor. Once the co-ordinates are obtained, then normalization is carried out based on the coordinate of the hip center for the first frame. All the coordinates are in the 3D space. The total dataset is divided into two parts for training and testing purposes. For the training procedure, 960 walking data sets are processed and the average walking data set is examined for each person. Remaining 320 walking data sets for each person are considered for testing purpose. The recognition procedure comprised with mean and standard deviation parameters with accuracy rate of 92.4786%.

Synergism of Firefly Algorithm and Q-Learning for Robot Arm Path Planning

Abstract Over the past few decades, Firefly Algorithm (FA) has attracted the attention of many researchers by virtue of its capability of solving complex real-world optimization problems. The only factor restricting the efficiency of this FA algorithm is the need of having balanced exploration and exploitation while searching for the global optima in the search-space. This balance can be established by tuning the two inherent control parameters of FA. One is the randomization parameter and another is light absorption coefficient, over iterations, either experimentally or by an automatic adaptive strategy. This paper aims at the later by proposing an improvised FA which involves the Q-learning framework within itself. In this proposed Q-learning induced FA (QFA), the optimal parameter values for each firefly of a population are learnt by the Q-learning strategy during the learning phase and applied thereafter during execution. The proposed algorithm has been simulated on fifteen benchmark functions suggested in the CEC 2015 competition. In addition, the proposed algorithms superiority is tested by conducting the Friedman test, Iman–Davenport and Bonferroni Dunn test. Moreover, its suitability for application in real-world constrained environments has been examined by employing the algorithm in the path planning of a robotic manipulator amidst various obstacles. To avoid obstacles one mechanism is designed for the robot-arm. The results, obtained from both simulation and real-world experiment, confirm the superiority of the proposed QFA over other contender algorithms in terms of solution quality as well as run-time complexity.

Real-time surface material identification using infrared sensor to control speed of an arduino based car like mobile robot

Recently autonomous cars are being in demand to launch into the market for safety and luxury. Autonomous car is an interesting area of research for Engineers and Researchers. Running an automated car on the road in real-time requires several factors to consider. Among them detecting the nature and type of the road is one major aspect. For different types of road surfaces, the control of speed, acceleration, break etc are required to adjust in a proper manner. This paper gives a very cost effective and efficient way of designing a surface identifier for fully autonomous cars. Using the surface identification module the car can identify the surface just in front of it and accordingly it can adjust its speed to move safely.

Type 2 fuzzy induced person identification using Kinect sensor

Automatic person recognition problem draws significant popularity in the last decade in the field of human-robot interaction. This paper introduces a novel approach to identify a person automatically whom the robot has already met, based on its walking pattern as gait is a unique characteristic for every individual. Here, the Kinect sensor is used to record the gait pattern of a person by storing 20 3-D joint coordinates in each time stamps. The features like joint angle and joint length are obtained from each complete walk cycle. Among all these features, most significant features are selected using principal component analysis. Later, these features are fuzzified constructing a Gaussian membership function with the mean and standard deviation of each feature at different gait cycle. An Interval Type-2 membership is constructed with all these membership values for a particular feature in different trials. 10 walking data set of 10 subjects are processed here. Now, when any person out of these 10 persons is walking in front of Kinect, features are calculated. But as more than one feature value for a particular feature (each feature corresponds to each gait cycle in a complete walking task) is obtained, mean of all these values for a particular feature is considered as measurement point. Defuzzification is done using t-norm and average operators. The person corresponding to highest defuzzified value is considered as the unknown person. The classification accuracy is 89.667%. The proposed method is also compared with few existing person identification techniques and the results obtained prove the superiority of the proposed algorithm.

Online Template Matching Using Fuzzy Moment Descriptor

In this paper a real-time template matching algorithm has been developed using Fuzzy (Type-1 Fuzzy Logic) approach. The Fuzzy membership-distance products, called Fuzzy moment descriptors are estimated using three common image features, namely edge, shade and mixed range. Fuzzy moment description matching is used instead of existing matching algorithms to reduce real-time template matching time. In the proposed matching technique template matching is done invariant to size, rotation and color of the image. For real time application the same algorithm is applied on an Arduino based mobile robot having wireless camera. Camera fetches frames online and sends them to a remote computer for template matching with already stored template in the database using MATLAB. The remote computer sends computed steering and motor signals to the mobile robot wirelessly, to maintain mobility of the robot. As a result, the mobile robot follows a particular object using proposed template matching algorithm in real time.