In this paper, we propose an MPC-based motion planning algorithm, including a decision-making module, an obstacle-constraints generating module, and an MPC-based planning module. The designed decision module effectively distinguishes between structured and unstructured roads and processes them separately, so that the algorithm is more robust in different environments. Besides, the movement of obstacles is considered in the decision-making and obstacle constraints generating module. By processing obstacles with lateral and longitudinal speed separately, obstacle avoidance can be done in scenarios with moving obstacles, including moving obstacles crossing the road. Instead of treating the vehicle as a mass point, we explicitly consider the geometric constraints by modeling the vehicle as three intersecting circles when generating obstacle constraints. This ensures that the vehicle is collision-free in motion planning, especially when the vehicle turns. For non-convex obstacle constraints, we propose an algorithm that generates up to two alternative linear constraints to convexify the obstacle constraints for improving computational efficiency. In MPC, we consider the vehicle kino-dynamic constraints and two generated linear constraints. Therefore, the proposed method can achieve better real-time performance and can be applied to more complicated traffic scenarios with moving obstacles. Simulation results in three different scenarios show that motion planning can achieve satisfactory performance in both structured and unstructured roads with moving obstacles.

Applying intelligent robot arms in dynamic uncertain environments (i.e., flexible production lines) remains challenging, which requires efficient algorithms for real time trajectory generation. The motion planning problem for robot trajectory generation is highly nonlinear and nonconvex, which usually comes with collision avoidance constraints, robot kinematics and dynamics constraints, and task constraints (e.g., following a Cartesian trajectory defined on a surface and maintain the contact). The nonlinear and nonconvex planning problem is computationally expensive to solve, which limits the application of robot arms in the real world. In this paper, for redundant robot arm planning problems with complex constraints, we present a motion planning method using iterative convex optimization that can efficiently handle the constraints and generate optimal trajectories in real time. The proposed planner guarantees the satisfaction of the contact-rich task constraints and avoids collision in confined environments. Extensive experiments on trajectory generation for weld grinding are performed to demonstrate the effectiveness of the proposed method and its applicability in advanced robotic manufacturing.

This paper addresses the autonomous robot ergodicity problem for efficient environment exploration. The spatial distribution as a reference distribution is given by a mixture of Gaussian and the mass generation of the robot is assumed to be skinny Gaussian. The main problem to solve is then to find out proper timing for the robot to visit as well as leave each component-wise Gaussian for the purpose of achieving the ergodicity. The novelty of the proposed method is that no approximation is required for the developed method. Given the definition of the ergodic function, a convergence condition is derived based on the timing analysis. Also, a formal algorithm to achieve the ergodicity is provided. To support the validity of the proposed algorithm, simulation results are provided.

Path-to-trajectory conversion problem for car-like autonomous ground vehicles has been studied in various ways. It is challenging to generate a trajectory which is dynamically feasible for the vehicle and comfortable for the passengers. An important practical concern of low computational costs makes the problem even more difficult. In this work, a path-to-trajectory converter is developed using a novel receding-horizon type suboptimal algorithm. By transforming the dynamic constraints to a longitudinal velocity limit function in the velocity-acceleration phase plane, a time-sub-optimal trajectory satisfying the dynamic constraints and the initial boundary condition is generated by computing the maximum constant acceleration in the down-range horizon. The portion of the trajectory approaching the end of the path is generated in reverse time. As illustrated by some simulation results and validation on a full-scale Kia Soul EV, the proposed path-to-trajectory conversion algorithm is able to account for dynamic constraints of the vehicle and guarantees passenger comfort.

Autonomous vehicles (AVs) need to interact with other traffic participants who can be either cooperative or aggressive, attentive or inattentive. Such different characteristics can lead to quite different interactive behaviors. Hence, to achieve safe and efficient autonomous driving, AVs need to be aware of such uncertainties when they plan their own behaviors. In this paper, we formulate such a behavior planning problem as a partially observable Markov Decision Process (POMDP) where the cooperativeness of other traffic participants is treated as an unobservable state. Under different cooperativeness levels, we learn the human behavior models from real traffic data via the principle of maximum likelihood. Based on that, the POMDP problem is solved by Monte-Carlo Tree Search. We verify the proposed algorithm in both simulations and real traffic data on a lane change scenario, and the results show that the proposed algorithm can successfully finish the lane changes without collisions.

This paper presents a strategy to integrate the planning and control for autonomous vehicles. The aim of this work is to provide a method that can yield controller feasible reference paths, i.e. paths that are not only dynamically feasible but are feasible under the action of a low-level feedback controller. The method is designed to find a control feasible parameterization of a collision-free path provided by a path generation scheme, e.g. rapidly-exploring random trees or one of its many variants. This parameterization is found such that the vehicle under the action of the low-level controller will be able to follow that path within a specified tolerance. The design is based on a feedback strategy with nested MPCs for planning and control. The results presented here are preliminary but hint at the benefits of such a strategy and suggest avenues for future work.

We investigate the dynamics of a pair of spinning spheres or microrotors in a fluid at low Reynolds numbers. These microrotors are each approximated by a rotlet, a fundamental singularity of the Stokes equation. Singularities of Stokes flows serve as useful theoretical models for microswimmers and micro-robots. Rotlet models of microswimmers have received less attention since a rotlet cannot generate translation by itself if the only control input is the rate of spin or strength of the rotlet. However a pair of rotlets can interact and execute net motion. In an unbounded domain of fluid, the positions of a pair of rotlets are not fully controllable due to the existence of an invariant. However, in a confined domain, we show that the positions of the pair of spheres are small time locally controllable. We show how control inputs can be constructed based on combinations of Lie brackets to move the rotlets from one point to another in the domain.

There has been a recent increase in research related to supernumerary robotic arms. A challenge with supernumerary robotic arms is how to operate them effectively. One solution is to use the foot to teleoperate the arm. That frees the person to use their arms for other tasks. However, unlike hand interfaces, it is not known how to create effective foot control for robotic teleoperation. This paper presents an experiment to compare position and rate control methods for foot interfaces. A foot interface is presented that can be used for both position and rate control. A human subject experiment uses 2D positioning tasks to evaluate the effectiveness of each control method. These same tasks are tested with a hand interface to provide a baseline for comparison. Results show that, similar to the hand, position control performs faster than rate control when using the foot.

In this research, a scooping motion generation method is proposed to scoop the semi-liquid objects from different containers automatically for meal support purpose. A spoon equipped robot arm is used. Based on the pre-measured shape of the containers, the robot arm can move the spoon to trace the inner surface of containers continuously. We also control the rotation of the spoon to scoop more semi-liquid object every time by imitating human’s scooping motion. A scraping motion is also generated as the auxiliary operation to gather the remaining semi-liquid object, which can realize an increase in the scooping amount. In the experiment, we tested the generated scooping motion for two containers and four type of semi-liquid objects. The scooped amount and the scooping times are measured and compared. The result shows that about 85.9% object on average could be scooped out.

In this paper, we address a visibility-based target tracking game for the scenario when the environment contains a circular obstacle. The game is originally formulated in four dimensions, but due to the symmetry of the environment, the dimension of the state space can be reduced to three. In the reduced state space, we formulate a game of degree. We use Isaacs techniques to obtain the locally optimal solution of the players. These solutions are extended back in the state space keeping in mind the geometry of the environment to obtain complete solution in the solution. Finally, we construct vector fields whose integral curves are the optimal solutions in the state space.

Citrus greening (Huanglongbing) has caused significant financial loss in many states of the United States and worldwide. In recent years, a heat therapy method has been investigated to prolong the production period of the diseased citrus trees after infection. One crucial step in this heat treatment process is to precisely align the truck with the diseased tree to reduce the operation time during deployment of the treatment tent. In this study, a binocular vision system is used to detect the position of the diseased tree relative to the truck, and a direct based path planning method is used to generate a nominal, optimal path for the truck to follow. A driver’s eye perception model is derived, simulating the distortion due to the human eye’s perception of objects on the computer screen, which will be used in the truck controller. A linear quadratic controller is then designed to compensate for the error coming from the eye perception mismatches and sensor and actuator noise. The studied human augmented driving control system can significantly reduce the operation time as the driver doesn’t have to constantly get out of the cab to check the truck-tree alignment. Simulation results show the effectiveness of the proposed system.

Kinematic motion planning using geometric mechanics tends to prescribe a trajectory in a parameterization of a shape space and determine its displacement in a position space. Often this trajectory is called a gait. Previous works assumed that the shape space is Euclidean when often it is not, either because the robotic joints can spin around forever (i.e., has an 𝕊 1 configuration space component, or its parameterization has an 𝕊 1 dimension). Consider a shape space that is a torus; gaits that “wrap” around the full range of a shape variable and return to its starting configuration are valid gaits in the shape space yet appear as line segments in the parameterization. Since such a gait does not form a closed loop in the parameterization, existing geometric mechanics methods cannot properly consider them. By explicitly analyzing the topology of the underlying shape space, we derive geometric tools to consider systems with toroidal and cylindrical shape spaces.

For a lately constructed disease detection field robot, the segregation of unhealthy leaves from strawberry plants is a major task. In field operations, the picking mechanism is actuated via three previously derived inverse kinematic algorithms and their performances are compared. Due to the high risk of rapid and unexpected deviation from the target position under field circumstances, some compensation is considered necessary. For this purpose, an image-based visual servoing method via the camera-in-hand configuration is activated when the end-effector is nearby to the target leaf subsequent to performing the inverse kinematics algorithms. In this study, a bio-inspired trajectory optimization method is proposed for visual servoing and the method is constructed based on a prey-predator relationship observed in nature (“motion camouflage”). In this biological phenomenon, the predator constructs its path in a certain subspace while catching the prey. The proposed algorithm is tested both in simulations and in hardware experiments.

In this paper, we propose a trust-based runtime verification (RV) framework for deploying multiple quad-rotors with a human-in-the-loop (HIL). By bringing together approaches from runtime verification, trust-based decision-making, human-robot interaction (HRI), and hybrid systems, we develop a unified framework that is capable of integrating human cognitive skills with autonomous capabilities of multi-robot systems to improve system performance and maximize the intuitiveness of the human-robot-interaction. On top of the RV framework, we utilize a probabilistic trust inference model as the key component in forming the HRI, designed to maintain the system performance. A violation avoidance controller is designed to account for the unexpected/unmodeled environment behaviors e.g. collision with static/moving obstacles. We also use the automata theoretic approaches to generate motion plans for the quad-rotors working in a partially-known environment by automatic synthesis of controllers enforcing specifications given in temporal logic languages. Finally, we illustrated the effectiveness of this framework as well as its feasibility through a simulated case study.

Artificial Potential Field (APF) theory is a unique branch of robotic path planning, which could be capable of handling the need for high dimensional robotic obstacle avoidance. However, APF theories have general performance issues which often make them undesirable in application. This research analyzes the Secant Approach; an algorithm developed to follow the APF style of path planning, but which has guaranteed convergence and obstacle avoidance properties in n-dimensional space. Using a unique potential function, the Secant Approach can guarantee a global minimum at the target while provably eliminating local minimums at other locations. Also, a control scheme has been developed which has guaranteed convergence properties. The Secant Approach is therefore capable of guiding various forms of robotic applications to target positions in n-dimensional space, making the theory a powerful path planning tool. This analysis examines the structure of the Secant Approach and extends the theory to include variable radius, solid obstacles.

In recent years, self-balancing personal transportation devices have gained significant popularity, the most popular ones being the “Hoverboard” systems. These systems utilize the dynamics of an inverted pendulum to create stable lateral motion. In this paper, the dynamic behavior of a hoverboard system with an attached flexible beam is investigated. By introducing a flexible beam, the vibrational characteristics of the system created by both the rider and the environment can be measured and accounted for. The beam is a continuous system modeled as an n degree of freedom (DOF) inverted pendulum. The resulting system becomes an n+2 DOF (n DOF for the beam, one DOF for the rotation of the beam about the wheel axis, and one DOF for the horizontal motion of the system). A mathematical model is developed to simulate the vibrations of the beam when excited by a piezoelectric actuator at the base, and to simulate the horizontal motion necessary to balance the beam as it is excited.

A majority of the routing algorithms for unmanned aerial or ground vehicles rely on Global Positioning System (GPS) information for localization. However, disruption of GPS signals, by intention or otherwise, can render these algorithms ineffective. This article provides a way to address this issue by utilizing landmarks to aid localization in GPS-denied environments. Specifically, given a number of vehicles and a set of targets, we formulate a joint routing and landmark placement problem as a combinatorial optimization problem: to compute paths for the vehicles that traverse every target at least once, and to place landmarks to aid the vehicles in localization while each of them traverses its route, such that the sum of the traveling cost and the landmark placement cost is minimized. A mixed-integer linear program is presented, and a set of algorithms and heuristics are proposed for different approaches to address certain issues not covered by the linear program. The performance of each proposed algorithm is evaluated and compared through extensive computational and simulation results.

In recent years, gliding robotic fish have emerged as promising mobile platforms for underwater sensing and monitoring due to their notable energy efficiency and maneuverability. For sensing of aquatic environments, it is important to use efficient sampling strategies that incorporate previously observed data in deciding where to sample next so that the gained information is maximized. In this paper, we present an adaptive sampling strategy for mapping a scalar field in an underwater environment using a gliding robotic fish. An ergodic exploration framework is employed to compute optimal exploration trajectories. To effectively deal with the challenging complexity of finding optimum three-dimensional trajectories that are feasible for the gliding robotic fish, we propose a novel strategy that combines a unicycle model-based 2D trajectory optimization with spiral-enabled water column sampling. Gaussian process (GP) regression is used to infer the field values at unsampled locations, and to update a map of expected information density (EID) in the environment. The outputs of GP regression are then fed back to the ergodic exploration engine for trajectory optimization. We validate the proposed approach with simulation results and compare its performance with a uniform sampling grid.

In this work, we address the problem of deploying a multi-robot system for row crop phenotyping. We propose a sampling-based navigation algorithm for the team of robots to estimate the underlying spatial distribution in a field. We use Gaussian Process Models to predict the distribution of a scalar function in a field, and choose Mutual Information as a metric for selecting the future samples. With a row crop structure, we present a collision-free assignment and scheduling algorithm for the robots to reach the goal positions which minimizes the total traveling distance. The effectiveness of the proposed algorithm is demonstrated through simulations.

Rapidly Exploring Random Trees (RRTs) have gained significant attention due to provable properties such as completeness and asymptotic optimality. However, offline methods are only useful when the entire problem landscape is known a priori . Furthermore, many real world applications have problem scopes that are orders of magnitude larger than typical mazes and bug traps that require large numbers of samples to match typical sample densities, resulting in high computational effort for reasonably low-cost trajectories. In this paper we propose an online trajectory optimization algorithm for uncertain large environments using RRTs, which we call Locally Adaptive Rapidly Exploring Random Tree (LARRT). This is achieved through two main contributions. We use an adaptive local sampling region and adaptive sampling scheme which depend on states of the dynamic system and observations of obstacles. We also propose a localized approach to planning and re-planning through fixing the root node to the current vehicle state and adding tree update functions. LARRT is designed to leverage local problem scope to reduce computational complexity and obtain a total lower-cost solution compared to a classical RRT of a similar number of nodes. Using this technique we can ensure that popular variants of RRT will remain online even for prohibitively large planning problems by transforming a large trajectory optimization approach to one that resembles receding horizon optimization. Finally, we demonstrate our approach in simulation and discuss various algorithmic trade-offs of the proposed approach.