Ahmed Hussain Qureshi

Intelligent bidirectional rapidly-exploring random trees for optimal motion planning in complex cluttered environments

Abstract The sampling-based motion planning algorithm known as Rapidly-exploring Random Trees (RRT) has gained the attention of many researchers due to their computational efficiency and effectiveness. Recently, a variant of RRT called RRT* has been proposed that ensures asymptotic optimality. Subsequently its bidirectional version has also been introduced in the literature known as Bidirectional-RRT* (B-RRT*). We introduce a new variant called Intelligent Bidirectional-RRT* (IB-RRT*) which is an improved variant of the optimal RRT* and bidirectional version of RRT* (B-RRT*) algorithms and is specially designed for complex cluttered environments. IB-RRT* utilizes the bidirectional trees approach and introduces intelligent sample insertion heuristic for fast convergence to the optimal path solution using uniform sampling heuristics. The proposed algorithm is evaluated theoretically and experimental results are presented that compares IB-RRT* with RRT* and B-RRT*. Moreover, experimental results demonstrate the superior efficiency of IB-RRT* in comparison with RRT* and B-RRT in complex cluttered environments.

Potential guided directional-RRT* for accelerated motion planning in cluttered environments

Recently proposed Rapidly Exploring Random Tree Star (RRT*) algorithm which is an extension of Rapidly Exploring Random Tree (RRT) provides collision free asymptotically optimal path regardless of obstacles geometry in a given environment. However, the drawback of this technique is a slow processing rate. This paper presents our proposed Potential Guided Directional-RRT* which addresses this problem and provides accelerated processing rate by incorporating Artificial Potential Fields Algorithm into RRT*. Artificial Potential Field algorithm directs the random samples toward the goal which leads to an increase in the speed of RRT*. We have presented simulation results of our technique and their comparison with results of RRT* under different environmental conditions to demonstrate apace execution rate of our novel idea.

Robot gains social intelligence through multimodal deep reinforcement learning

For robots to coexist with humans in a social world like ours, it is crucial that they possess human-like social interaction skills. Programming a robot to possess such skills is a challenging task. In this paper, we propose a Multimodal Deep Q-Network (MDQN) to enable a robot to learn human-like interaction skills through a trial and error method. This paper aims to develop a robot that gathers data during its interaction with a human, and learns human interaction behavior from the high dimensional sensory information using end-to-end reinforcement learning. This paper demonstrates that the robot was able to learn basic interaction skills successfully, after 14 days of interacting with people.

Potential functions based sampling heuristic for optimal path planning

Rapidly-exploring Random Tree star (RRT*) is a recently proposed extension of Rapidly-exploring Random Tree (RRT) algorithm that provides a collision-free, asymptotically optimal path regardless of obstacles geometry in a given environment. However, one of the limitation in the RRT* algorithm is slow convergence to optimal path solution. As a result it consumes high memory as well as time due to the large number of iterations utilised in achieving optimal path solution. To overcome these limitations, we propose the potential function based-RRT* that incorporates the artificial potential field algorithm in RRT*. The proposed algorithm allows a considerable decrease in the number of iterations and thus leads to more efficient memory utilization and an accelerated convergence rate. In order to illustrate the usefulness of the proposed algorithm in terms of space execution and convergence rate, this paper presents rigorous simulation based comparisons between the proposed techniques and RRT* under different environmental conditions. Moreover, both algorithms are also tested and compared under non-holonomic differential constraints.

Adaptive Potential guided directional-RRT

The Rapidly Exploring Random Tree Star (RRT*) is an extension of the Rapidly Exploring Random Tree path finding algorithm. RRT* guarantees an optimal, collision free path solution but is limited by slow convergence rates and inefficient memory utilization. This paper presents APGD-RRT*, a variant of RRT* which utilizes Artificial Potential Fields to improve RRT* performance, providing relatively better convergence rates. Simulation results under different environments between the proposed APGD-RRT* and RRT* algorithms demonstrate this marked improvement under various test environments.

Human tracking by a mobile robot using 3D features

Detection and Tracking of human being is a very important problem in Computer Vision. Human robot interaction is a very essential need for service robots where robots are required to detect and track human beings in order to provide the required service. In this paper we present an improved novel approach for tracking a target person in crowded environment. We used multi-sensor data fusion approach by combining the data of stereo camera and laser rangefinder (LRF) to perform human tracking. The system gathers the features of human upper body, face and legs in the target person selection phase and then the robot will start following the target person. Camera is used for upper body and face detection while laser rangefinder is used for gathering legs data. Template matching and triangulation is done in order to get the dimensions of upper body and face. Target person tracking is done using Cam shift tracker. Thus our method presents a novel approach that uses all these techniques to track a target person in a crowded environment.

Show, attend and interact: Perceivable human-robot social interaction through neural attention Q-network

For a safe, natural and effective human-robot social interaction, it is essential to develop a system that allows a robot to demonstrate the perceivable responsive behaviors to complex human behaviors. We introduce the Multimodal Deep Attention Recurrent Q-Network using which the robot exhibits human-like social interaction skills after 14 days of interacting with people in an uncontrolled real world. Each and every day during the 14 days, the system gathered robot interaction experiences with people through a hit-and-trial method and then trained the MDARQN on these experiences using end-to-end reinforcement learning approach. The results of interaction based learning indicate that the robot has learned to respond to complex human behaviors in a perceivable and socially acceptable manner.

Triangular geometry based optimal motion planning using RRT*-motion planner

RRT* is a recent and improved variant of the RRT path finding algorithm. While RRT concentrates on simply finding an initial obstacle-free path, RRT* guarantees eventual convergence to an optimum, collision-free path for any given geometrical environment. On the other hand, the main limitations of RRT* include its slow processing rate and high memory utilization due to the large number of iterations required to achieve optimal path solution. This paper presents Triangular Geometerised-RRT* (TG-RRT*) which incorporates Triangular geometrical methods in the RRT* algorithm and improves its processing time by decreasing the number of iterations required for optimal solution. Simulation results under different environments demonstrate an improved convergence rate of TG-RRT*, in comparison with RRT*.

Triangular Geometrized Sampling Heuristics for Fast Optimal Motion Planning

Rapidly-exploring Random Tree (RRT)-based algorithms have become increasingly popular due to their lower computational complexity as compared with other path planning algorithms. The recently presented RRT* motion planning algorithm improves upon the original RRT algorithm by providing optimal path solutions. While RRT determines an initial collision-free path fairly quickly, RRT* guarantees almost certain convergence to an optimal, obstacle-free path from the start to the goal points for any given geometrical environment. However, the main limitations of RRT* include its slow processing rate and high memory consumption, due to the large number of iterations required for calculating the optimal path. In order to overcome these limitations, we present another improvement, i.e, the Triangular Geometerized-RRT* (TG-RRT*) algorithm, which utilizes triangular geometrical methods to improve the performance of the RRT* algorithm in terms of the processing time and a decreased number of iterations required for an optimal path solution. Simulations comparing the performance results of the improved TG-RRT* with RRT* are presented to demonstrate the overall improvement in performance and optimal path detection.

Intrinsically motivated reinforcement learning for human-robot interaction in the real-world

For a natural social human-robot interaction, it is essential for a robot to learn the human-like social skills. However, learning such skills is notoriously hard due to the limited availability of direct instructions from people to teach a robot. In this paper, we propose an intrinsically motivated reinforcement learning framework in which an agent gets the intrinsic motivation-based rewards through the action-conditional predictive model. By using the proposed method, the robot learned the social skills from the human-robot interaction experiences gathered in the real uncontrolled environments. The results indicate that the robot not only acquired human-like social skills but also took more human-like decisions, on a test dataset, than a robot which received direct rewards for the task achievement.