1 An oscillator-based smooth real-time estimate of gait phase for wearable robotics Tingfang Yan‧Andrea Parri‧Virginia Ruiz Garate‧Marco Cempini‧Renaud Ronsse‧Nicola Vitiello Abstract This paper presents a novel methodology for estimating the gait phase of human walking through a simple sensory apparatus. Three subsystems are combined: a primary phase estimator based on adaptive oscillators, a desired gait event detector and a phase error compensator. The estimated gait phase is expected to linearly increase from 0 to 2π rad in one stride and remain continuous also when transiting to the next stride. We designed two experimental scenarios to validate this gait phase estimator, namely treadmill walking at different speeds and free walking. In the case of treadmill walking, the maximum phase error at the desired gait events was found to be 0.155 rad, and the maximum phase difference between the end of the previous stride and beginning of the current stride was 0.020 rad. In the free walking trials, phase error at the desired gait event was never larger than 0.278 rad. Our algorithm outperformed against two other benchmarked methods. The good performance of our gait phase estimator could provide consistent and finely tuned assistance for an exoskeleton designed to augment the mobility of patients. Keywords Real-time gait phase estimate, adaptive oscillators, phase error learning, wearable robotics. 1 Introduction Autonomous mobility is essential for daily-life activities. However, the prevalence of lower-limb mobility impairments is observed among people at different ages, especially the elderly (Iezzoni et al. 2001). It is estimated that up to 35% of people over 70 encounter problems in walking. This percentage increases to almost 60% at 80-84 years of age (Verghese et al. 2006). Gait impairments are thus widespread among elderly people and can lead to both physical and cognitive barriers: slower walking, health-risk situations, such as stumbling, and thus limited social participation or depression. In an era where various innovative technologies have been developed for daily-life services, wearable robotics designed to assist people with gait impairments is attracting great attention, for example lower-limb exoskeletons/orthoses. The assistive functions of a lower-limb exoskeleton rely on the capabilities to: (i) “decode” the user’s intended movement; and (ii) detect the real-time phase of this intended movement (Pons 2008). In particular, the success of these interpretations is fundamental for the assistive device to consistently act with natural gait biomechanics, thus restoring more functional and energy-efficient locomotion. In this study, we focus on the development of the second capability, i.e. acquiring an accurate and continuous gait phase for the locomotion task. For the purpose of a friendly and comfortable human-robot interface, the phase estimator is expected to be independent from a complex sensory apparatus. In the literature, different methods of estimating the gait phase for powered lower-limb exoskeletons can be tracked (Yan et al. 2015). A classic approach is designed to firstly identify the human gait as a series of time-discrete states and then implement the encoded assistive function phase by phase. The discrete states can be easily recognized with wearable sensors, such as the application on the exoskeleton “BLEEX” (Kazerooni et al. 2006) and the transfemoral prosthesis in (Ambrozic et al. 2014). This approach – despite being relatively easy to be implemented through finite-state classifiers – has some limitations. First of all, the identification of each state only relies on the data fed back by a wearable sensory apparatus. Thus it is prone to induce a delay. Furthermore, the discrete states limit the possibility to smoothly provide and finely tune assistance as a continuous function of the locomotion phase. In order to overcome these limitations, an alternative strategy to detect the gait phase in real time is developed by exploiting the biomechanical information received from the muscle level. This method could anticipate the movement detection before the actual kinematic change: for instance, the utilization by HAL (Kawamoto et al. 2010) and Michigan ankle foot orthosis (Kao et al. 2010). Nevertheless, a trade-off is that the robot-user interface turns out to be more complex, mainly due to the T. Yan is the corresponding author, The BioRobotics Institute, Scuola Superiore Sant'Anna, viale Rinaldo Piaggio 34, 56025 Pontedera (PI), Italy. Tel.: +39 3334900778, Fax: +39 050883101. E-mail: t.yan@sssup.it; A. Parri (an.parri@sssup.it); M. Cempini (m.cempini@sssup.it). The BioRobotics Institute, Scuola Superiore Sant'Anna, viale Rinaldo Piaggio 34, 56025 Pontedera (PI), Italy. V. Ruiz Garate (virginia.ruizgarate@uclouvain.be); R. Ronsse (renaud.ronsse@uclouvain.be). Louvain Bionics and Institute of Mechanics, Materials, and Civil Engineering, Université catholique de Louvain, Place du Levant, 2 bte L5.04.02 B-1348 Louvain-la-Neuve, Belgium. N. Vitiello (n.vitiello@sssup.it) The BioRobotics Institute, Scuola Superiore Sant'Anna, viale Rinaldo Piaggio 34, 56025 Pontedera (PI), Italy. Don Carlo Gnocchi Foundation, Florence, Via di Scandicci, 265, Florence, Italy. The first two authors contributed equally to this work.