Table 1.
Comparison between our review and the current reviews in the literature.
| Topic | Description | |
|---|---|---|
| Vanderborght et al. (2013) and Wolf et al. (2016) | Variable impedance actuators | Realize Variable Impedance Control (VIC) in hardware with dedicated elastic elementsa. They reviewed all possibilities to create variable stiffness actuators and all main factors that influence the most common approaches. |
| Calanca et al. (2015) | Compliance control | Reviewed impedance and admittance controllers for both stiff and soft joint robots. |
| Keemink et al. (2018) | Admittance control | Reviewed admittance controllers with a specific focus on human–robot interaction. |
| Song et al. (2019) | All above | This review compared hardware- and software-based approaches, and main technical developments about impedance control including hybrid impedance, force-tracking, and adaptive methods. However, learning algorithms and VIC methods are mentioned in two small subsections. |
| Our review | VIC, VIL, and VILC | This review departs from impedance control approaches to focus on learning and learning control approaches used to implement variable impedance behaviors. We analyze the advantages and disadvantages of traditional approaches based on control and recent frameworks that integrate learning techniques. Therefore, our review has a potential impact on both the control and the learning communities. |
SEA are implemented by introducing intentional elasticity between the motor actuator and the load for robust force control, which subsequently improves safety during the interaction of the robot with the environment (Pratt and Williamson, 1995).
SEA are out of the scope of this paper, however, interested readers can refer to Calanca et al. (2017), which summarizes the common controller architectures for SEA. SEA framework has been used in many applications, mainly in human–robot interaction scenarios. For instance, Yu et al. (2013, 2015) designed compliant actuators for gait rehabilitation robot and validated its controller in order to provide safety and stability during the interaction. Li et al. (2017) proposed a multi-modal controller for exoskeleton rehabilitation robots, driven by SEA, that guarantee the stability of the system. For more recent approaches, please refer to Haninger et al. (2020) and Kim et al. (2020).
On the other hand, for recent achievements in VIC in soft robots, please refer to (Ataka et al., 2020; Gandarias et al., 2020; Li et al., 2020; Sozer et al., 2020; Zhong et al., 2020).