This paper presents a differential-flatness-based integrated point-to-point trajectory planning and control method for a class of nonholonomic wheeled mobile manipulator (WMM). We demonstrate that its kinematic model possesses a feedback-linearizable description due to the flatness property, which allows for full-state controllability. Trajectory planning can then be simplified and achieved by polynomial fitting method in the flat output space to satisfy the terminal conditions, while control design reduces to a pole-placement problem for a linear system. The method is then deployed on our custom-constructed WMM hardware to evaluate its effectiveness and to highlight various aspects of the hardware implementation.

Prototyping and validating hardware–software components, sub-systems and systems within the intelligent transportation system-of-systems framework requires a modular yet flexible and open-access ecosystem. This work presents our attempt to develop such a comprehensive research and education ecosystem, called AutoDRIVE, for synergistically prototyping, simulating and deploying cyber-physical solutions pertaining to autonomous driving as well as smart city management. AutoDRIVE features both software as well as hardware-in-the-loop testing interfaces with openly accessible scaled vehicle and infrastructure components. The ecosystem is compatible with a variety of development frameworks, and supports both single- and multi-agent paradigms through local as well as distributed computing. Most critically, AutoDRIVE is intended to be modularly expandable to explore emergent technologies, and this work highlights various complementary features and capabilities of the proposed ecosystem by demonstrating four such deployment use-cases: (i) autonomous parking using probabilistic robotics approach for mapping, localization, path-planning and control; (ii) behavioral cloning using computer vision and deep imitation learning; (iii) intersection traversal using vehicle-to-vehicle communication and deep reinforcement learning; and (iv) smart city management using vehicle-to-infrastructure communication and internet-of-things.

Assisted motor therapies play a critical role in enhancing functional musculoskeletal recovery and neurological rehabilitation. Our long-term goal is to assist and automate the performance of repetitive motor-therapy of the human lower limbs. Hence, in this paper, we examine the viability of a light-weight and reconfigurable hybrid (articulated-multibody and cable) robotic system for assisting lower-extremity rehabilitation and analyze its performance. A hybrid cable-actuated articulated-multibody system is formed when multiple cables are attached from a ground-frame to various locations on an articulated-linkage-based orthosis. Our efforts initially focus on developing an analysis and simulation framework for the kinematics and dynamics of the cable-driven lower limb orthosis. A Monte Carlo approach is employed to select configuration parameters including cuff sizes, cuff locations, and the position of fixed winches. The desired motions for the rehabilitative exercises are prescribed based upon motion patterns from a normative subject cohort. We examine the viability of using two controllers—a joint-space feedback-linearized PD controller and a task-space force-control strategy—to realize trajectory- and path-tracking of the desired motions within a simulation environment. In particular, we examine performance in terms of (i) coordinated control of the redundant system; (ii) reducing internal stresses within the lower-extremity joints; and (iii) continued satisfaction of the unilateral cable-tension constraints throughout the workspace.

Cable-driven parallel manipulators (CDPM) potentially offer many advantages over serial manipulators, including greater structural rigidity, greater accuracy, and higher payload-to-weight ratios. However, CDPMs possess limited moment resisting/exerting capabilities and relatively small orientation workspaces. Various methods have been contemplated for overcoming these limitations, each with its own advantages and disadvantages. The focus of this paper is on one such method: the addition of base mobility to the system. Such base mobility gives rise to kinematic redundancy, which needs to be resolved carefully in order to control the system. However, this redundancy can also be exploited in order to optimize some secondary criteria, e.g., maximizing the size and quality of the wrench-closure workspace with the addition of base mobility. In this work, the quality of the wrench-closure workspace is examined using a tension-factor index. Two planar mobile base configurations are investigated, and their results are compared with a traditional fixed-base system. In the rectangular configuration, each base is constrained to move along its own linear rail, with each rail forming right angles with the two adjacent rails. In the circular configuration, the bases are constrained to move along one circular rail. While a rectangular configuration enhances the size and quality of the orientation workspace in a particular rotational direction, the circular configuration allows for the platform to obtain any position and orientation within the boundary of the base circle. Furthermore, if the bases are configured in such a way that the cables are fully symmetric with respect to the platform, a maximum possible tension-factor of one is guaranteed. This fully symmetric configuration is shown to offer a variety of additional advantages: it eliminates the need to perform computationally expensive nonlinear optimization by providing a closed-form solution to the inverse kinematics problem, and it results in a convergence between kinematic singularities and wrench-closure singularities of the system. Finally, we discuss a particular limitation of this fully symmetric configuration: the inability of the cables to obtain an even tension distribution in a loaded configuration. For this reason, it may be useful to relax the fully symmetric cable requirement in order to yield reasonable tensions of equal magnitude.

Knowledge distillation (KD) has become a widely used technique for model compression and knowledge transfer. We find that the standard KD method performs the knowledge alignment on an individual sample indirectly via class prototypes and neglects the structural knowledge between different samples, namely, knowledge correlation. Although recent contrastive learning-based distillation methods can be decomposed into knowledge alignment and correlation, their correlation objectives undesirably push apart representations of samples from the same class, leading to inferior distillation results. To improve the distillation performance, in this work, we propose a novel knowledge correlation objective and introduce the dual-level knowledge distillation (DLKD), which explicitly combines knowledge alignment and correlation together instead of using one single contrastive objective. We show that both knowledge alignment and correlation are necessary to improve the distillation performance. In particular, knowledge correlation can serve as an effective regularization to learn generalized representations. The proposed DLKD is task-agnostic and model-agnostic, and enables effective knowledge transfer from supervised or self-supervised pretrained teachers to students. Experiments show that DLKD outperforms other state-of-the-art methods on a large number of experimental settings including: 1) pretraining strategies; 2) network architectures; 3) datasets; and 4) tasks.