Recent studies have reported on the remarkable ability of bioinspired soft robots to exhibit dexterous and contact-friendly motions. However, for these robots with deformable bodies, it is extremely challenging to achieve precise and robust control when undergoing uncertain forces and contact in the environment. In this work, we take a first step to address this issue for slender pneumatic soft robots by proposing a comprehensive modeling and control framework. Our framework employs a fully parametrized model that accurately describes both robot configurations and distributed forces using Hermite interpolation. Leveraging this model, we further establish an estimation algorithm that can infer complete robot configurations and distributed external forces from limited motion data, enabling perception of contact locations and forces. Integrating this model and estimator, our control framework achieves precise robot motion control under diverse forces, with the average trajectory tracking error within 0.3 mm. It also detects and adapts to uncertain contact, demonstrated in tests of automatic obstacle avoidance and precise grasping. This framework holds promise for various applications such as environmental exploration and safe manipulation, where compliant interaction with the environment is required.

Parallel robots with rigid transmission mechanisms have been widely developed to improve the speed, precision, and load capability. However, it is still challenging in promoting soft parallel robots due to the difficulty in accurate kinematic modeling of soft continuum links. In this article, we present a general framework on the design, kinematic modeling, model-based characterization, and control for a class of soft parallel robots. The designed soft parallel robot consists of three fiber-reinforced soft pneumatic actuators, a base stage, and an output stage. With the introduction of the mathematical toolkit of the absolute nodal coordinate formulation, we develop a continuum-based model to describe and parameterize both the global complex configuration and the local large deformation of the soft parallel robot. In this sense, the mappings among the defined kinematic spaces of the robot can be characterized through force analysis. Based on the developed model, we next analyze the robot’s workspace and stiffness with different design parameters, which are also verified by a set of experiments. Finally, we establish a model-based trajectory tracking controller for the soft parallel robot. The experimental results demonstrate that with the feedforward controller, the end effector of the soft parallel robot can well follow the desired trajectories under different output velocities, where the average positioning error is about 2.6–3.9% of the maximum length of the workspace.