Search

Underactuated robots often require involved routines for trajectory planning due to their complex dynamics. Flapping-wing aerial vehicles have unsteady aerodynamics and periodic gaits that complicate the planning procedure. In this paper, we improve upon existing methods for flight planning by introducing a two-stage optimization routine to plan flapping flight trajectories. The first stage solves a trajectory optimization problem with a data-driven fixed-wing approximation model trained with experimental flight data. The solution to this is used as the initial guess for a second stage optimization using a flapping-wing model trained with the same flight data. We demonstrate the effectiveness of this approach with a bat robot in both simulation and experimental flight results. The speed of convergence, the dependency on the initial guess, and the quality of the solution are improved, and the robot is able to track the optimized trajectory of a dive maneuver.

In this paper, we present self-supervised shared latent embedding (S3LE), a data-driven motion retargeting method that enables the generation of natural motions in humanoid robots from motion capture data or RGB videos. While it requires paired data consisting of human poses and their corresponding robot configurations, it significantly alleviates the necessity of time-consuming data-collection via novel paired data generating processes. Our self-supervised learning procedure consists of two steps: automatically generating paired data to bootstrap the motion retargeting, and learning a projection-invariant mapping to handle the different expressivity of humans and humanoid robots. Furthermore, our method guarantees that the generated robot pose is collision-free and satisfies position limits by utilizing nonparametric regression in the shared latent space. We demonstrate that our method can generate expressive robotic motions from both the CMU motion capture database and YouTube videos.

In this paper, we present the second version of a reconfigurable modular legged robot, Snapbot V2. The mechanical design of Snapbot V2 is enhanced for better dynamic performance and robust connection with modular legs. A motion generator for locomotion is developed to achieve various locomotion skills in one to six-leg configurations. The locomotion is tested on a multi-body dynamic simulation model and implemented on a physical robot as well. A visual detection is implemented with a camera module to recognize the robot’s configuration. By detecting the particular color of the parts at the leg module, the robot can recognize the number and location of the connected legs. Based on the recognized configuration, Snapbot V2 selects the proper locomotion style automatically.

We present a novel computational approach to design the robotic devices from high-level motion specifications. Our computational system uses a library of modular components—actuators, mounting brackets, and connectors—to define the space of possible robot designs. The process of creating a new robot begins with a set of input trajectories that specify how its end effectors and/or body should move. By searching through the combinatorial set of possible arrangements of modular components, our method generates a functional, as-simple-as-possible robotic device that is capable of tracking the input motion trajectories. To significantly improve the efficiency of this discrete optimization process, we propose a novel heuristic that guides the search for appropriate designs. Briefly, our heuristic function estimates how much an intermediate robot design needs to change before it becomes able to execute the target motion trajectories. We demonstrate the effectiveness of our computational design method by automatically creating a variety of robotic manipulators and legged robots. To generate these results, we define our own robotic kit that includes off-the-shelf actuators and 3-D printable connectors. We validate our results by fabricating two robotic devices designed with our method.

In the DARPA Robotics Challenge DRC, participating human-robot teams were required to integrate mobility, manipulation, perception, and operator interfaces to complete a simulated disaster mission. We describe our approach using the humanoid robot Atlas Unplugged developed by Boston Dynamics. We focus on our approach, results, and lessons learned from the DRC Finals to demonstrate our strategy, including extensive operator practice, explicit monitoring for robot errors, adding additional sensing, and enabling the operator to control and monitor the robot at varying degrees of abstraction. Our safety-first strategy worked: we avoided falling, and remote operators could safely recover from difficult situations. We were the only team in the DRC Finals that attempted all tasks, scored points 14/16, did not require physical human intervention a reset, and did not fall in the two missions during the two days of tests. We also had the most consistent pair of runs.

Imitation of the upper body motions of human demonstrators or animation characters to human-shaped robots is studied in this paper. We present a pipeline for motion retargeting by transferring the joints of interest (JOI) of source motions to the target humanoid robot. To this end, we deploy an optimization-based motion retargeting method utilizing link length modifications of the source skeleton and a task (Cartesian) space fine-tuning of JOI motion descriptors. To evaluate the effectiveness of the proposed pipeline, we use two different 3-D motion datasets from three human demonstrators and an Ogre animation character, Bork, and successfully transfer the motions to four different humanoid robots: DARwIn-OP, COmpliant HuMANoid Platform (COMAN), THORMANG, and Atlas. Furthermore, COMAN and THORMANG are actually controlled to show that the proposed method can be deployed to physical robots.

In this paper, we propose a trajectory-based reinforcement learning method named deep latent policy gradient (DLPG) for learning locomotion skills. We define the policy function as a probability distribution over trajectories and train the policy using a deep latent variable model to achieve sample efficient skill learning. We first evaluate the sample efficiency of DLPG compared to the state-of-the-art reinforcement learning methods in simulated environments. Then, we apply the proposed method to a four-legged walking robot named Snapbot to learn three basic locomotion skills of turn left, go straight, and turn right. We demonstrate that, by properly designing two reward functions for curriculum learning, Snapbot successfully learns the desired locomotion skills with moderate sample complexity.

In this paper, we present an airbag fabrication machine and its process to make airbags which can be utilized for human-machine interaction. The proposed machine uses a CNC-style gantry with XYZ axes to seal layers of thermoplastic sheets in intricate patterns. A custom designed ball transfer is mounted on this machine to control temperature to press, heat, and fuse the thermoplastic sheets. We also implemented a path planning algorithm for heat sealing based on user's drawing. Using the developed machine and process, we demonstrate fabricated airbags can be pressed more than our target pressure and durable to be used in multiple applications over time.

In this paper, we present an automated learning environment for developing control policies directly on the hardware of a modular legged robot. This environment facilitates the reinforcement learning process by computing the rewards using a vision-based tracking system and relocating the robot to the initial position using a resetting mechanism. We employ two state-of-the-art deep reinforcement learning (DRL) algorithms, Trust Region Policy Optimization (TRPO) and Deep Deterministic Policy Gradient (DDPG), to train neural network policies for simple rowing and crawling motions. Using the developed environment, we demonstrate both learning algorithms can effectively learn policies for simple locomotion skills on highly stochastic hardware and environments. We further expedite learning by transferring policies learned on a single legged configuration to multi-legged ones.

This paper presents a 6 DOF leg mechanism for a humanoid robot. The proposed Hybrid Leg is designed to combine serial and parallel mechanisms and consists of a pair of twin 3 DOF serial chains in parallel. A 5-bar-linkage mechanism is implemented to the serial mechanism to generate 2 DOF motion regarding hip and knee pitch rotation. The hardware prototype is designed by matching the kinematic specification of a commercial robot's leg to compare the proposed mechanism with a conventional serial leg. We derive the analytical expressions of its forward and inverse kinematics. End-effector workspaces are shown with plots and inverse dynamics analysis of Hybrid Leg and serial leg with a given walking gait trajectory is presented. Hardware experiment is conducted with a prototype to verify the simulated workspace and trajectory tracking performance.

The DARPA Robotics Challenge (DRC) required participating human-robot teams to integrate mobility, manipulation, perception and operator interfaces to complete a simulated disaster mission. We describe our approach to the development of manipulation and locomotion capabilities for the humanoid robot atlas unplugged developed by Boston Dynamics. We focus on our approach, results and lessons learned from the DRC Finals to demonstrate our strategy including extensive operator practice, explicit monitoring for robot errors, adding additional sensing, and enabling the operator to control and monitor the robot at varying degrees of abstraction. Our safety-first strategy worked: we avoided falling and remote operators could safely recover from difficult situations. We were the only team in the DRC Finals that attempted all tasks, scored points (14/16), did not require physical human intervention (a reset), and did not fall in the two missions during the two days of tests. We also had the most consistent pair of runs. We ranked 3rd out of 23 teams when the scores from two official runs were averaged.

This paper summarizes observations and lessons learned by the WPI-CMU team and self-reports made by many of the DARPA Robotics Challenge teams on what happened at the DARPA Robotics Challenge Finals. Major conclusions are: (1) Reducing operator errors is the most cost effective way to improve robot performance. Methods include operator training and practice, and software safeguards to detect and prevent operator errors. (2) Super-human sensing is another way to greatly improve robot performance. To some extent this matches what happened in the DARPA autonomous driving challenges, in which improved sensing was the key to improved performance. (3) Paradigm shifts are needed in academic robotics, such as emphasizing designing robust behaviors, systems design including what seem like unimportant issues such as thermal management, and consistent real world results rather than videos of the rare successes.

Current and previous single-legged hopping robots are energetically tethered and lack portability. Here, we present the design and control of an untethered, energetically autonomous single-legged hopping robot. The thrust-producing mechanism of the robot’s leg is an actuated prismatic joint, called a linear elastic actuator in parallel (LEAP). The LEAP mechanism comprises a voice coil actuator in parallel with two compression springs, which gives our robot passive compliance. An actuated gimbal hip joint is realized by two standard servomotors. To control the robot, we adapt Raibert’s hopping controller, and find we can maintain balance roughly in-place for up to approx. 7 s (19 hops) while continuously hopping.

To enable robust dynamic walking on the Atlas robot, we extend our previous work by adding a receding-horizon component. The new controller consists of three hierarchies: a center of mass (CoM) trajectory planner that follows a sequence of desired foot steps, a receding-horizon controller that optimizes the next foot placement to minimize future CoM tracking errors, and an inverse dynamics based full body controller that generates instantaneous joint commands to track these motions while obeying physical constraints. An approximate value function is generated by the CoM planner, and is used to guide the foot placement and inverse dynamics optimizations. The proposed controller is implemented and tested on the Atlas robot. It is capable of walking with strong external perturbations such as recovering from large pushes and traversing unstructured terrain.

The design of legged robots is often inspired by animals evolved to excel at different tasks. However, while mimicking morphological features seen in nature can be very powerful, robots may need to perform motor tasks that their living counterparts do not. In the absence of designs that can be mimicked, an alternative is to resort to mathematical models that allow the relationship between a robot's form and function to be explored. In this paper, we propose such a model to co-design the motion and leg configurations of a robot such that a measure of performance is optimized. The framework begins by planning trajectories for a simplified model consisting of the center of mass and feet. The framework then optimizes the length of each leg link while solving for associated full-body motions. Our model was successfully used to find optimized designs for legged robots performing tasks that include jumping, walking, and climbing up a step. Although our results are preliminary and our analysis makes a number of simplifying assumptions, our findings indicate that the cost function, the sum of squared joint torques over the duration of a task, varies substantially as the design parameters change.

We have developed a toy sized humanoid robot with soft air-filled modules on its links which sense contact and protect the robot and any interacting humans from damaging collisions. This robot, meant for robust physical interaction, is required to endure contact with children in the form of hugs and other playful interactions. It is therefore necessary to quantify the forces exerted during these interactions so that robots can be designed to both withstand these forces, as well as interact safely and intuitively in these situations. To quantify the range of forces exerted by children when performing both soft and strong hugs, we conducted a study in which 28 children (11 boys, 17 girls) between 4 and 10 years old hugged a pressure sensing doll while the pressure was recorded. We found a child's maximum expected hugging force (2.623 psi for our setup) during free play. The data gathered in this study will guide the further development of our physically interactive robot.

Effective teleoperation requires real-time control of a remote robotic system. In this work, we develop a controller for realizing smooth and accurate motion of a robotic head with application to a teleoperation system for the Furhat robot head [1], which we call TeleFurhat. The controller uses the head motion of an operator measured by a Microsoft Kinect 2 sensor as reference and applies a processing framework to condition and render the motion on the robot head. The processing framework includes a pre-filter based on a moving average filter, a neural network-based model for improving the accuracy of the raw pose measurements of Kinect, and a constrained-state Kalman filter that uses a minimum jerk model to smooth motion trajectories and limit the magnitude of changes in position, velocity, and acceleration. Our results demonstrate that the robot can reproduce the human head motion in real time with a latency of approximately 100 to 170 ms while operating within its physical limits. Furthermore, viewers prefer our new method over rendering the raw pose data from Kinect.

Among legged robots, hopping and running robots are useful because they can traverse terrain at high speeds and are a benchmark platform for locomotion actuators; if an actuator can power a hopping robot, it can power a walking robot. We aim to create a hopping mechanism for a small-scale, one-legged, untethered hopping robot. A parallel-elastic actuator is an efficient way to do this, and enables the actuator to directly inject energy into the spring, but requires a high-speed, low-inertia actuator. Voice coil actuators are electrically-powered direct-drive translational motors that have very low moving inertia, low friction, can produce force at high speeds, and have a linear force output. These qualities make them ideal candidate motors for a linear elastic actuator in parallel (“LEAP”). Here, we derive an electromechanical model of the LEAP mechanism, develop a simple bang-bang hopping controller, and simulate hopping with a range of spring parameters to find an optimal spring stiffness that maximizes hopping height. We detail our implemented design, and characterize its performance through a series of experiments. We test our robot with different spring stiffnesses, and demonstrate hopping at a maximum steady-state of 3.5 cm ground-clearance (approx. 20% leg length). Our results suggest that the LEAP mechanism may serve the weight-bearing functions of a robot leg.

This paper presents an effective and systematic trajectory generation method, together with a control method for enabling a biped robot to walk upstairs. The COG (center of gravity) trajectory is generated by the VHIPM (virtual height inverted pendulum mode) for the horizontal motion and by a 6th order polynomial for the vertical motion; an ankle compliance control (ACC) is also added into the robot control. The proposed methods are evaluated by simulations as well as being implemented in a robot for the performance verification. The results show that the proposed methods can generate stable motions when walking upstairs, and these can significantly reduce the zero moment point (ZMP) errors compared with other methods, enabling the robot to walk up steeper stairs.

To physically interact with children, a robot should be safe and durable. We have developed a small toy sized robot with a soft skin that is robust to playful physical interaction. The upper body, including the arms, pelvis, chest and back has soft, 3D printed air-filled modules connected to pressure sensors to sense contact and provide protection to the child and robot while interacting. These soft skin modules cover the underlying actuators and the rigid 3D printed frame with printed bearings. In this paper, we present the design process of these modules and demonstrate their efficacy by implementing a "grab and move" user interface for posing the robot link by link.

We describe the design and hardware implementation of our walking and manipulation controllers that are based on a cascade of online optimizations. A virtual force acting at the robot's center of mass (CoM) is estimated and used to compensated for modeling errors of the CoM and unplanned external forces. The proposed controllers have been implemented on the Atlas robot, a full size humanoid robot built by Boston Dynamics, and used in the DARPA Robotics Challenge Finals, which consisted of a wide variety of locomotion and manipulation tasks.

Our goal is to bring animation characters to life in the real world. We present a bipedal robot that looks like and walks like an animation character. We start from animation data of a character walking. We develop a bipedal robot which corresponds to lower part of the character following its kinematic structure. The links are 3D printed and the joints are actuated by servo motors. Using trajectory optimization, we generate an open-loop walking trajectory that mimics the character's walking motion by modifying the motion such that the Zero Moment Point stays in the contact convex hull. The walking is tested on the developed hardware system.

To achieve dynamic balancing and natural walking for a bipedal robot we propose a novel force-based control framework. Given 6-dimensional pose vector representing robot's posture and attitude, desired force and moment in the task space are computed. To generate the force and moment as desired, we propose the use of virtual gravity compensation (VGC), essentially a dynamic controller that outputs joint torques. By using the VGC-based balancing controller, the robot can maintain a desired pose stably even on a tilting plate. We also propose to extend the VGC-based balancing controller to implement a walking algorithm that controls the desired pose in terms of capture point using a finite state machine. The control algorithm was tested with torque-controlled humanoid platforms developed by our group to demonstrate robust and natural gaits under various walking environments. The robot walked robustly on irregular surfaces and recovered from external pushes. The robot also exhibited natural walking motions such as pendulum-like leg swings and heel-to-toe transitions, a characteristic feature of human gait, all without explicitly designating joint angle trajectories.

This paper presents a framework to generate dynamic walking for biped robots. A set of self-stable gait primitives is first constructed. It is done by 1) representing parametric gait primitives, 2) utilizing state-dependent torque control, and 3) doing numerical optimization that takes into account the complex multi-body dynamics with frictional contact forces. Dynamic walking to follow the arbitrary path including a curve is then generated online via sequentially composing primitive motions. Results show that dynamic gaits are humanlike and efficient compared to the conventional knee bent walkers. Our proposed method is applied to a torque-controlled, human-sized biped robot platform, ‘Roboray’ which is cable-driven partially for joint compliance. Following a discussion on robot design and control, experimental results are also reported.

This paper gives an overview of the development of a novel biped walking machine for a humanoid robot, Roboray. This lower-limb robot is designed as an experimental system for studying biped locomotion based on force and torque controlled joints. The robot has 13 actuated DOF and torque sensors are integrated at all the joints except the waist joint. We designed a new tendon type joint modules as a pitch joint drive module, which is highly back-drivable and elastic. We also built a decentralized control system using the small controller boards named Smart Driver. The forward walking experiment with this lower limbs was conducted to test the mechanical structure and control system.

Bipedal walking robots are inherently hybrid systems due to their intermittent, switching dynamics resulting from the impact between the robot foot and the ground as the robot foot lands on the ground. It is well known that stable (passive) limit cycles for the biped robots can be induced on shallow slopes without actuation. Recently the studies in passive dynamic walking have considered the robots with knee and point or curved feet. In this paper, we study the passive dynamic walking for biped robots with knee and fixed flat feet, which includes heel and toe rocking motions and the effect of foot length on the passive limit cycles. We derive the dynamic equations of motion for this model. We show by simulation that the proposed robot model can walk down a slope passively and also verify the stability of this walking by calculating the eigenvalues of the Jacobian of the Poincare map. By using a numerical search method, we find the initial conditions of the stable limit cycles for various slope angles and foot lengths.

We propose a control framework for dynamic bipedal locomotion with compliant joints. A novel 3D dynamic walking is achieved by utilizing natural dynamics of the system. It is done by 1) driving robot joints directly with the posture-based state machine and 2) controlling tendon-driven compliant actuators. To enlarge gait's basin attraction for stable walking, we also adaptively plan step-to-step motion and compensate stance/swing motion. Final joint input is described by a superposition of state machine control torques and compensation torques of balancers. Various walking styles are easily generated by composing straight and turning gait-primitives and such walking is effectively able to adapt on various environments. Our proposed method is applied to a torque controlled robot platform, Roboray. Experimental results show that gaits are able to traverse inclined and rough terrains with bounded variations, and the result gaits are human-like comparing the conventional knee bent walkers.

In this paper, we introduce a new algorithm called Multi-Population Genetic Algorithm (MPGA) as an effi- cient optimization technique for highly nonlinear problems. Our MPGA is a parallel implementation of a GA and shows its robustness for our problem domain (Task Based Design; TBD) where we design a manipulator which is best suited for a given task. The MPGA has the same number of optimization fuac- tions as the number of task points and maintains almost con- stant complexity and does not depend on the number of task points. In addition, we develop a framework called Progressive Design, in which we design progressively based on coarse-6ne approach.