Collaboration between humans and robots is an important aspect of Industry 4.0. It can be improved by incorporating human-like characteristics into robot motion planning. It is assumed that humans move optimal with respect to a certain objective or cost function. To find this function, also for a robot, we use an inverse optimal control approach identifying what linear weighted combination of physically interpretable cost functions best mimics human point-to-point motions. A bi-level optimization is used, where the upper level compares the optimal robot result of the lower level with human reference motions. Two depth cameras are combined in a setup to record these reference motions. The resulting weighted cost functions are then used to generate new motions for a seven degrees of freedom robot arm. The resulting optimized motions are compared to standard robot motions based on linear interpolation in joint or task space. The comparison is performed by means of a small experiment where preliminary observations show that humans experience these motions as more anthropomorphic and feel at least equally comfortable and safe compared to existing motion planning strategies.

@en

This article describes Team Delft's robot winning the Amazon Robotics Challenge 2016. The competition involves automating pick and place operations in semi-structured environments, specifically the shelves in an Amazon warehouse.
Team Delft's entry demonstrated that current robot technology can already address most of the challenges in product handling: object recognition, grasping, motion, or task planning; under broad yet bounded conditions. The system combines an industrial robot arm, 3D cameras and a custom gripper. The robot's software is based on the Robot Operating System to implement solutions based on deep learning and other state-of-the-art artificial intelligence techniques, and to integrate them with off-the-shelf components.
From the experience developing the robotic system it was concluded that: 1) the specific task conditions should guide the selection of the solution for each capability required, 2) understanding the characteristics of the individual solutions and the assumptions they embed is critical to integrate a performing system from them, and 3) this characterization can be based on `levels of robot automation'. This paper proposes automation levels based on the usage of information at
design or runtime to drive the robot's behaviour, and uses them to discuss Team Delft's design solution and the lessons learned from this robot development experience.
@en

Robotic systems are the workhorses in practically all automated applications. Manufacturing industries, warehouses, elderly care, disaster rescue and (unfortunately) warfare are example applications where human life has benefited from robotics. By precisely planning controlling their motions via computer programs, real world tasks can be performed with high levels of accuracy and repeatability. Devising methods and algorithms that generate such motions by a) correctly reporting and finding the desired motion if it exists and b) doing so as fast as possible, has constituted the field of robot motion planning research over the last four decades. In recent years, the Industry 4.0 initiative has provided a promising avenue for further advances in industrial automation. Modular, quickly reconfigurable and versatile robotic systems that safely collaborate with humans hold the key to future industrial automation. This is a challenging endeavor from an industrial and an academic perspective and inspires the work in this thesis. In alignment with these perspectives, this thesis is presented in two parts. In the first part, we propose methods and frameworks to effectively utilize open source implementations of configuration space planners to realize flexible and robust solutions for bin picking. To this end, three results are presented: a tool to automatically tune parameters of path planning algorithm implementations, a world championship winning solution for industrial bin picking and a reactive collision avoidance framework for collaborative robotic applications. Configuration space planners are extremely popular due to their solution speeds of about a tenth of a second for planning problems in 7-8 dimensions. However, a primary limitation of configuration space planners is that their planning solutions do not account for the physical laws governing the movement of robots. Consequently, the possibilities of generating versatile and dynamically feasible motions are highly curtailed. This limitation can be addressed by planning in the state space. However, sampling-based planning in state space is computationally intensive and challenging to realize in practice. This challenge inspires the second part of this thesis. Here, the goal is to answer the question: Is it possible to achieve planning speeds in state space that are comparable to planning speeds in the configuration space? We pursue this goal by considering the Rapidly exploring Random Tree (RRT) planner in state space to plan a swing-up motion for a simple pendulum. Here, we propose two contributions that alleviate the computational demands of two critical steps in the RRT planner. We present a framework to approximate the distance (pseudo) metric and the steering function in state space using supervised learning tools. Together, a speed up of about 4 orders of magnitude is achieved relative to numerically solving for these two critical steps. However, reaching planning times equivalent to or better than what is achievable in configuration space still remains an elusive goal. Nevertheless, the achieved results serve as encouraging signs to pursue further research in this direction. @en

Sampling-based kinodynamic planners, such as Rapidly-exploring Random Trees (RRTs), pose two fundamental challenges: computing a reliable (pseudo-)metric for the distance between two randomly sampled nodes, and computing a steering input to connect the nodes. The core of these challenges is a Two Point Boundary Value Problem, which is known to be NP-hard. Recently, the distance metric has been approximated using supervised learning, reducing computation time drastically. The previous work on such learning RRTs use direct optimal control to generate the data for supervised learning. This paper proposes to use indirect optimal control instead, because it provides two benefits: it reduces the computational effort to generate the data, and it provides a low dimensional parametrization of the action space. The latter allows us to learn both the distance metric and the steering input to connect two nodes. This eliminates the need for a local planner in learning RRTs. Experimental results on a pendulum swing up show 10-fold speed-up in both the offline data generation and the online planning time, leading to at least a 10-fold speed-up in the overall planning time.

@en

A large number of novel path planning methods for a wide range of problems have been described in literature over the past few decades. These algorithms can often be configured using a set of parameters that greatly influence their performance. In a typical use case, these parameters are only very slightly tuned or even left untouched. Systematic approaches to tune parameters of path planning algorithms have been largely unexplored. At the same time, there is a rising interest in the planning and robotics communities regarding the real world application of these theoretically developed and simulation-tested planning algorithms. In this work, we propose the use of Sequential Model-based Algorithm Configuration (SMAC) tools to address these concerns. We show that it is possible to improve the performance of a planning algorithm for a specific problem without the need of in-depth knowledge of the algorithm itself. We compare five planners that see a lot of practical usage on three typical industrial pick-and-place tasks to demonstrate the effectiveness of the method.@en

This paper describes Team Delft’s robot, which won the Amazon Picking Challenge 2016, including both the Picking and the Stowing competitions. The goal of the challenge is to automate pick and place operations in unstructured environments, specifically the shelves in an Amazon warehouse. Team Delft’s robot is based on an industrial robot arm, 3D cameras and a customized gripper. The robot’s software uses ROS to integrate off-the-shelf components and modules developed specifically for the competition, implementing Deep Learning and other AI techniques for object recognition and pose estimation, grasp planning and motion planning. This paper describes the main components in the system, and discusses its performance and results at the Amazon Picking Challenge 2016 finals.

@en