Unsupervised deep learning based ego motion estimation with a downward facing camera

Knowing the robot's pose is a crucial prerequisite for mobile robot tasks such as collision avoidance or autonomous navigation. Using powerful predictive models to estimate transformations for visual odometry via downward facing cameras is an understudied area of research. This work proposes a novel approach based on deep learning for estimating ego motion with a downward looking camera. The network can be trained completely unsupervised and is not restricted to a specific motion model. We propose two neural network architectures based on the Early Fusion and Slow Fusion design principle: “EarlyBird” and “SlowBird”. Both networks share a Spatial Transformer layer for image warping and are trained with a modified structural similarity index (SSIM) loss function. Experiments carried out in simulation and for a real world differential drive robot show similar and partially better results of our proposed deep learning based approaches compared to a state-of-the-art method based on fast Fourier transformation.

Novel surface optimization for trajectory reconstruction in industrial robot tasks

This study presents a novel algorithm implementation that optimizes manually recorded toolpaths with the use of a 3D-workpiece model to reduce manual error induced. The novel algorithm has three steps: workpiece declaration, manual toolpath declaration, and toolpath optimization using steepest descent algorithm. Steepest descent finds the surface route wherein the manually recorded toolpaths traverse over a 3D-workpiece surface. The optimized toolpaths were simulated and tested with an industrial robot showing minimal error compared to the desired optimized toolpaths. The results obtained from the presented implementation on three different trajectories demonstrate that the proposed methodology can reduce the manual error induced using as a reference the CAD-workpiece surface.

Development and validation of an end-effector for mitigation of collisions

In robotics the risk of collisions is present both in industrial applications and in remote handling. If a collision occurs, the impact may damage both the robot and external equipment, which may result in successive imprecise robot tasks or line stops, reducing robot efficiency. As a result, appropriate collision avoidance algorithms should be used or, if it is not possible, the robot must be able to react to impacts reducing the contact forces. For this purpose, this paper focuses on the development of a special end-effector that can withstand impacts. It is able to protect the robot from impulsive forces caused by collisions of the end-effector, but it has no effect on possible collisions between the links and obstacles. The novel end-effector is based on a bi-stable mechanism that decouples the dynamics of the end-effector from the dynamics of the robot. The intrinsically non-linear behavior of the endeffector is investigated with the aid of numerical simulations. The effect of design parameters and operating conditions are analyzed and the interaction between the functioning of the bi-stable mechanism and the control system is studied. In particular, the effect of the mechanism in different scenarios characterized by different robot velocities is shown. Results of numerical simulations assess the validity of the proposed end-effector, which can lead to large reductions in impact forces. Numerical results are validated by means of specific laboratory tests.

Development of an End-Effector for Mitigation of Collisions

In robotics the risk of collisions is present both in industrial applications and in remote handling. If a collision occurs, the impact may damage both the robot and external equipment, which may result in successive imprecise robot tasks or line stops, reducing robot efficiency. As a result, appropriate collision avoidance algorithms should be used or, if it is not possible, the robot must be able to react to impacts reducing the contact forces. For this purpose, this paper focuses on the development of a special end-effector that can withstand impacts and is able to protect the robot from impulsive forces. The novel end-effector is based on a bi-stable mechanism that decouples the dynamics of the end-effector from the dynamics of the robot. The intrinsically non-linear behavior of the end-effector is investigated with the aid of numerical simulations. The effect of design parameters and the operating conditions are analyzed and the interaction between the functioning of the bi-stable mechanism and the control system is studied. In particular, the effect of the mechanism in different scenarios characterized by different robot velocities is shown. Results of numerical simulations assess the validity of the proposed end-effector, which can lead to large reductions in impact forces.

Robot coalition formation against time-extended multi-robot tasks

PurposeMulti-robot coalition formation (MRCF) refers to the formation of robot coalitions against complex tasks requiring multiple robots for execution. Situations, where the robots have to participate in multiple coalitions over time due to a large number of tasks, are called Time-extended MRCF. While being NP-hard, time-extended MRCF also holds the possibility of resource deadlocks due to any cyclic hold-and-wait conditions among the coalitions. Existing schemes compromise on solution quality to form workable, deadlock-free coalitions through instantaneous or incremental allocations.Design/methodology/approachThis paper presents an evolutionary algorithm (EA)-based task allocation framework for improved, deadlock-free solutions against time-extended MRCF. The framework simultaneously allocates multiple tasks, allowing the robots to participate in multiple coalitions within their schedule. A directed acyclic graph–based representation of robot plans is used for deadlock detection and avoidance.FindingsAllowing the robots to participate in multiple coalitions within their schedule, significantly improves the allocation quality. The improved allocation quality of the EA is validated against two auction schemes inspired by the literature.Originality/valueTo the best of the author's knowledge, this is the first framework which simultaneously considers multiple MR tasks for deadlock-free allocation while allowing the robots to participate in multiple coalitions within their plans.

Planning and Reinforcement Learning for General-Purpose Service Robots

Despite recent progress in AI and robotics research, especially learned robot skills, there remain significant challenges in building robust, scalable, and general-purpose systems for service robots. This Ph.D. research aims to combine symbolic planning and reinforcement learning to reason about high-level robot tasks and adapt to the real world. We will introduce task planning algorithms that adapt to the environment and other agents, as well as reinforcement learning methods that are practical for service robot systems. Taken together, this work will make a significant step towards creating general-purpose service robots.

Exploring Non-Expert Robot Programming Through Crowdsourcing

A longstanding barrier to deploying robots in the real world is the ongoing need to author robot behavior. Remote data collection–particularly crowdsourcing—is increasingly receiving interest. In this paper, we make the argument to scale robot programming to the crowd and present an initial investigation of the feasibility of this proposed method. Using an off-the-shelf visual programming interface, non-experts created simple robot programs for two typical robot tasks (navigation and pick-and-place). Each needed four subtasks with an increasing number of programming statements (if statement, while loop, variables) for successful completion of the programs. Initial findings of an online study (N = 279) indicate that non-experts, after minimal instruction, were able to create simple programs using an off-the-shelf visual programming interface. We discuss our findings and identify future avenues for this line of research.