This paper presents clinical results from the use of monitor-nomad (MONIMAD), which is a reactive robotized interface for lower limb rehabilitation of patients suffering from cerebellar disease. The first problem to be addressed is the postural analysis of sit-to-stand motion. Experiments with healthy subjects were performed for this purpose. Analysis of external forces shows that sit-to-stand transfer can be subdivided into several phases: preacceleration, acceleration, start rising, and rising. Observation of center of pressure, ground forces, and horizontal components force on handles yields rules that identify the stability of the patient and adjust the robotic interface motion to the human voluntary movement. These rules are used in a fuzzy-based controller implementation. The controller is validated on experiments with diseased patients at Bellan Hospital, Paris, France.

Detecting human motion and predicting human intentions by analyzing body signals are challenging but fundamental steps for the implementation of applications presenting human–robot interaction in different contexts, such as robotic rehabilitation in clinical environments, or collaborative robots in industrial fields. Machine learning techniques (MLT) can face the limit of small data amounts, typical of this kind of applications. This paper studies the illustrative case of the reaching movement in 10 healthy subjects and 21 post-stroke patients, comparing the performance of linear discriminant analysis (LDA) and random forest (RF) in: (i) predicting the subject’s intention of moving towards a specific direction among a set of possible choices, (ii) detecting if the subject is moving according to a healthy or pathological pattern, and in the case of discriminating the damage location (left or right hemisphere). Data were captured with wearable electromagnetic sensors, and a sub-section of the acquired signals was required for the analyses. The possibility of detecting with which arm (left or right hand) the motion was performed, and the sensitivity of the MLT to variations in the length of the signal sub-section were also evaluated. LDA and RF prediction accuracies were compared: Accuracy improves when only healthy subjects or longer signals portions are considered up to 11% and at least 10%, respectively. RF reveals better estimation performance both as intention predictor (on average 59.91% versus the 62.19% of LDA), and health condition detector (over 90% in all the tests).

Robots following humans is an efficient and practical feature, particularly in the context of service robotics. However, the majority of existing research has focused on the robot following the human from behind, with relatively little attention given to the robot operating in front of the human. This following-in-front approach, where the robot remains within the user’s field of view, is more reassuring and facilitates further human-robot interaction. Unlike traditional following methods, frontal following requires awareness of the user’s orientation and intentions. New challenges will arise in developing a tracker that can accurately estimate a user’s pose based on a knee height 2D LiDAR, especially when the legs frequently occlude each other. There is also a need to ensure the safety of the user while addressing how the robot can keep up with the user in situations where it falls behind. Our contribution lies in proposing a novel 2D LiDAR-based frontal human following method that accommodates various motion patterns of the target user. Specifically, inspired by human walking gait, we develop an accurate and robust human pose tracker that takes into account leg occlusion and the data association problem. We build a large workspace velocity field that enables a holonomic mobile robot to follow and come gradually and safely in front of the user regardless of their relative position. We evaluate the performance of our approach through a rich set of experimental scenarios, and demonstrate its effectiveness in achieving reliable frontal human following. Our studies suggest this approach has potential applications in warehouses, industrial factories or for visually impaired persons.

This article presents the results of an experiment on physical Human-Human Interaction (pHHI), where human dyads cooperate on a one-dimensional comanipulative task in a novel lightweight teleoperation setup. The results of this experiment show that humans are able to handle asymmetrical information about the task and solve conflicts using only the kinesthetic channel. Data from the pHHI experiment is used to design a virtual partner that can perform the task alongside a human. The virtual partner behavior is based on the observation that initiative is highly correlated to decision-making in our pHHI negotiation scenario. The virtual agent is then evaluated in a physical Human-Robot Interaction (pHRI) experiment. The results of the second experiment show that the virtual partner is able to perform the task without compromising the performances of the dyad and that a similar role distribution is observed in human-human and human-robot dyads. Moreover, the knowledge of the partner’s nature does not seem to influence the performances. The results obtained with the virtual partner are encouraging and could be used to design kinesthetic negotiation algorithms in pHRI settings. The main contributions of this article are: (1) supporting evidence of the possibility for humans to use the kinesthetic channel as mean of negotiation during physical Human-Human Interaction in the lightest known impedance negotiation tasks (with no virtual mass involved), (2) highlighting of the correlation between initiative and dyadic decision making, (3) design of a simple yet efficient virtual partner algorithm capable of realistic physical Human-Robot Interaction in the one-dimension tracking task used in the experimental setup. This design combines minimum jerk trajectory, switching role ability, decision criteria, and statistical parameters that are detailed in this article.