Nuwan Ganganath

A Constraint-Aware Heuristic Path Planner for Finding Energy-Efficient Paths on Uneven Terrains

Motions of mobile robots need to be optimized to minimize their energy consumption to ensure long periods of continuous operations. Shortest paths do not always guarantee the minimum energy consumption of mobile robots. Moreover, they are not always feasible due to climbing constraints of mobile robots, especially on steep terrains. We utilize a heuristic search algorithm to find energy-optimal paths on hilly terrains using an established energy-cost model for mobile robots. The terrains are represented using grid-based elevation maps. Similar to A*-like heuristic search algorithms, the energy-cost of traversing through a given location of the map depends on a heuristic energy-cost estimation from that particular location to the goal. Using zigzag-like path patterns, the proposed heuristic function can estimate heuristic energy-costs on steep terrains that cannot be estimated using traditional methods. We proved that the proposed heuristic energy-cost function is both admissible and consistent. Therefore, the proposed path planner can always find feasible energy-optimal paths on any given terrain without node revisits, provided that such paths exist. Results of tests on real-world terrain models presented in this paper demonstrate the promising computational performance of the proposed path planner in finding energy-efficient paths.

Data Clustering with Cluster Size Constraints Using a Modified K-Means Algorithm

Data clustering is a frequently used technique in finance, computer science, and engineering. In most of the applications, cluster sizes are either constrained to particular values or available as prior knowledge. Unfortunately, traditional clustering methods cannot impose constrains on cluster sizes. In this paper, we propose some vital modifications to the standard k-means algorithm such that it can incorporate size constraints for each cluster separately. The modified k-means algorithm can be used to obtain clusters in preferred sizes. A potential application would be obtaining clusters with equal cluster size. Moreover, the modified algorithm makes use of prior knowledge of the given data set for selectively initializing the cluster centroids which helps escaping from local minima. Simulation results on multidimensional data demonstrate that the k-means algorithm with the proposed modifications can fulfill cluster size constraints and lead to more accurate and robust results.

A 2-Dimensional ACO-Based Path Planner for Off-Line Robot Path Planning

Wireless sensor networks are usually deployed in scenarios that are too hostile for human personnel to perform maintenance tasks. Wireless sensor nodes usually exchange information in a multi-hop manner. Connectivity is crucial to the performance of a wireless sensor network. In case a network is partitioned due to node failures, it is possible to re-connect the fragments by setting up bridges using mobile platforms. Given the landscape of a terrain, the mobile platforms should be able reach the target position using a desirable path. In this paper, an off-line robot path planner is proposed to find desirable paths between arbitrary points in a given terrain. The proposed path planner is based on ACO algorithms. Unlike ordinary ACO algorithms, the proposed path planner provides its artificial ants with extra flexibility in making routing decisions. Simulation results show that such enhancement can greatly improve the qualities of the paths obtained. Performances of the proposed path planner can be further optimized by fine-tuning its parameters.

Trajectory planning for 3D printing: A revisit to traveling salesman problem

Three dimensional (3D) printing can be used to manufacture many different objects range from toys to hi-tech robot parts. This paper investigates 3D printer trajectory planning to improve the speed of the printing process. The printing speed mainly depends on the motion speed and path of the printing nozzle. We use triangular and trapezoidal velocity profiles to minimize the transition time between print segments. In this work, several algorithms that were originally proposed as solutions for conventional traveling salesman problem are modified to adapt to the new problem. The proposed modifications are designed to obtain time-efficient trajectories for the printing nozzle.

A Real-Time ASL Recognition System Using Leap Motion Sensors

It is always challenging for deaf and speech-impaired people to communicate with non-sign language users. A real-time sign language recognition system using 3D motion sensors could lower the aforementioned communication barrier. However, most existing gesture recognition systems are adopting a single sensor framework, whose performance is susceptible to occlusions. In this paper, we proposed a real-time multi-sensor recognition system for American sign language (ASL). Data collected from Leap Motion sensors are fused using multiple sensors data fusion (MSDF) and the recognition is performed using hidden Markov models (HMM). Experimental results demonstrate that the proposed system can deliver higher recognition accuracy over single-sensor systems. Due to its low implementation cost and higher accuracy, the proposed system can be widely deployed and bring conveniences to sign language users.

An ACO-based off-line path planner for nonholonomic mobile robots

The path planning is an important issue as it allows a robot to get from a point to another. Such a path between two points has to be optimized based on user defined requirements and environmental conditions. Most of the existing solutions to path planning problem assume robots to be holonomic. However, ordinary mobile robots are kinematically constrained in practice. A novel solution has been proposed for the path planning problem based on existing ant colony optimization methods which can be realized with practical mobile robots with the aforementioned constraints. Simulation results show the applicability of the proposed path planner to the nonholonomic mobile robots. Performance of the proposed algorithm has been compared with its preceding version. The performance of the proposed path planner may be further improved by fine-tuning its parameters.

Live demonstration: A HMM-based real-time sign language recognition system with multiple depth sensors

Automatic sign language recognition plays an important role in communications for sign language users. Most existing sign language recognition systems use single sensor input. However, such systems may fail to recognize hand gestures correctly due to occluded regions of hand gestures. In this work, we propose a novel system for real-time recognition of the digits in American Sign Language (ASL) [1]. The proposed system [2] utilizes two Leap Motion sensors [3] to capture hand gestures from different angles. Sensory data are preprocessed using a multi-sensor data fusion approach and ASL digits are recognized in real-time from the fused data using Hidden Markov models (HMM) [4]. Experimental results of the proposed sign language recognition system demonstrate its improved performance over single sensor systems. With a low implementation cost and a high recognition accuracy, the proposed system can be widely adopted in many real world applications and bring conveniences to world-wide ASL users.

Finding energy-efficient paths on uneven terrains

Mobile robots are increasingly getting popular in outdoor applications. Long period of continuous operations are common in such applications. Therefore, robot motions need to be optimized to minimize their energy consumption. Shortest paths do not always guarantee minimum energy consumptions of mobile robots. This paper proposes a novel algorithm to generate energy-efficient paths on uneven terrains using an established energycost model for mobile robots. Terrains are represented using grid based elevation maps. Similar to A* algorithm, the energy-cost of traversing through a particular gird depends on a heuristic energy-cost estimation from the current location to the goal. The proposed heuristic energy-cost function makes it possible to generate zigzag-like path patterns on steep hills under the power limitations of the robot. Therefore, the proposed method can find physically feasible energy-efficient paths on any given terrain, provided that such paths exist. Simulation results presented in this paper demonstrate the performance of the proposed algorithm on uneven terrains maps.

A 3D printing path optimizer based on Christofides algorithm

Rapid prototyping and product customization have become more convenient with the emergence of 3D printing technologies. In extrusion deposition based 3D printing, objects are built by connecting many lines of filament, layer by layer. The efficiency of the printing process can be improved by optimizing motion paths of the printing nozzle. In this paper, a 3D printing path optimizer based on Christofides algorithm is proposed. Experiment results show that the proposed optimizer can significantly reduce the length of motion paths compared to a nearest neighbor-based optimizer using in consumer 3D printers.

Rapid replanning of energy-efficient paths for navigation on uneven terrains

Mobile robots are often utilized in remote and hostile outdoor environments with uncertainties and unknown dangerous. The energy-efficient paths generated based on prior information can be impracticable due to the changes in the environment. Recently proposed Z* search algorithm is capable of finding physically feasible energy-efficient paths on uneven terrains. It can achieve the same accuracy as any brute force algorithm, but with a low computational complexity. However, neither Z* nor any other energy-efficient path planners can effectively handle path replanning triggered by environment changes such as emergence of obstacles. In order to fill this void, we propose a novel algorithm which can recompute optimal paths efficiently. Simulation results show that the proposed algorithm can find equally energy-efficient paths as Z* does, but at a considerably lower computational cost. Therefore, the proposed algorithm can be very useful in mobile robot navigation on uneven terrains with unknown obstacles.