Experimental Validation of Adaptive Control for a Shape Memory Alloy Actuated Lightweight Robotic Arm

This article presents the experimental validation of a Direct Adaptive Control for angular position regulation of a lightweight robotic arm. The robotic arm is single degree-of-freedom (DOF) system, actuated by two Shape Memory Alloy (SMA) wires. The proposed adaptive control is capable of adapting itself to the hysteretic behavior of SMA wires and update its behavior to deal with the changing parameters of the material over time. The closed-loop approach is tested experimentally showing its effectiveness to deal with the highly nonlinear dynamics of the SMA wires. These results are discussed and compared with a classical control approach. The updated design and hardware development and modeling of the robotic arm are shown. INTRODUCTION In recent years, Shape Memory Technology (SMT) has become a trend in the research for alternative actuation systems. The term SMT is used to describe the implementation of materials with Shape Memory Effect (SME). This effect is the property of materials to recover their original shape, after being deformed, upon external stimuli. These stimuli can be thermal, chemical, mechanical, among others. A common type of materials with SMA is the Shape Memory Alloys (SMA). These materials include a group of alloys (most commonly Nickel∗Address all correspondence to this author. Titanium [1]) which can be easily deformed at lower temperatures and then recover their original shape when subject to proper mechanical or thermal stimuli. This transformation occurs due to an inner shifting in the material’s crystalline structure. At lower temperatures, the material transforms into martensite, a highly malleable phase. When the SMA is subjected to the proper external stimuli, it transforms into austenite phase, a rigid cubic structure, allowing the material to recover its original shape [2]. The increased interest in the implementation and research of these materials is due to their multiple advantages over conventional actuators, advantages such as high force to mass ratio, corrosion resistance, biocompatibility, noiseless operation, among others. These advantages make SMA suitable for a wide range of applications, ranging from biomedical and dental implants to aerospace engineering. Some SMA applications that we can find in the literature are for example, in medical areas as intra-arterial supports [3], adaptive anklefoot orthoses [4], skeletal fixation devices [5] or orthodontic applications [6]. We can also find implementations in aerodynamics as an actuator for morphing segments on UAV [7] or wing shape control [8]. Many other specific purpose applications have been reported in the literature such as camera lens focus actuators [9], car mirror actuators [10] or SMA based motors [11]. Other examples of this are the multiple robotics implementations. For instance, the robotic arm developed by the author in [12] or the anthropomorphic robotic fingers proposed by [13] and [14]. Along with 1 Copyright c © 2018 by ASME these specific purpose applications some more general, advanced control oriented application have been developed as the ones presented in [15–18]. In spite of their multiple advantages, SMAs also entail different kind of challenges for mechanism design and the control of them. Most of the aforementioned applications, imply a complicated mechanical design or are oriented to micro-scale actuation. To solve this problem, a lightweight design for an SMA actuated robotic arm was proposed in [19]. This proposal, using 3D printing technology and lightweight materials, together with a relatively simple mechanical design, achieves a lightweight robotic arm suitable for implementation in mobile environments where the weight and size is a primary concern, as flying manipulation. An improved mechanical design of this proposal is introduced in the current work. In addition to the mechanical challenges, the nonlinear dynamics of the SMAs make them difficult to control and model. Among these nonlinear dynamics, we found principally a highly hysteretic behavior, phenomena as dead zone or superelasticity. For this reason, multiple control approaches have been developed an reported in the literature. From simple classical control techniques as PID or Variable Structure Control (VSC) [20], to intelligent control as fuzzy control approach [13] or neural networks [12]. A set of techniques proven to be suitable to face the highly nonlinear behavior of these materials are those of the adaptive control. Different adaptive techniques have been implemented for SMA control. These techniques can be broadly classify in Direct and Indirect adaptive control [15–17, 21, 22] and intelligent adaptive control. This latter class includes techniques combining adaptive and intelligent control, like the ones reported in [23–25]. These intelligent adaptive controls depend on the identification of multiple parameters, and their quality relies on the number of neurons and persistent excitation conditions. Contrary to these methods, we propose a Direct adaptive control method where only one parameter needs to be tuned in real-time for each SMA wire. SMA WIRE ACTUATED ROBOT ARM DESIGN This section presents the mechanical design of the proposed SMA wire actuated robotic arm. Figure 1 shows a Computer Aided Design (CAD) model of the lightweight robotic arm. The mechanism presented here is based on a previous work published in [26,27]. The proposed design is a lightweight robotic arm with one Degree of Freedom (DOF), actuated by 3 SMA wires. The custom design of the pieces, together with 3D printing technology and light fabrication materials, permit the construction of a mechanism with an approximated weight of 50 g and a range of angular displacement up to 90 degrees. The robotic arm presented in [26, 27] is originally based on the joint proposed by the authors in [20]. This design proposes two couplers (coupler-1 and coupler-2), each one actuated by an individual SMA wire (SMA-1 and SMA-2 respectively), and joined with a torsion spring in between the couplers. This configuration allows for faster response of the overall system by controlling the total stiffness of the joint, thus increasing or decreasing the transformation temperatures, which leads to a faster forward or backward transformation of the SMA-1. The stiffness adjustment is carried out by the SMA-2 while the angular position of the joint, and the end-effector’s position, depends directly on the SMA-1. The current work proposes a new mechanical design for the robotic arm’s prototype. A couple of groove bearings are implemented, substituting the previously proposed winding wheels (see [27]). This change allows lower overall friction, thus higher angular displacements while handling the SMA wires relative high temperatures adequately. Furthermore, an entirely new mechanical design for the gripper and its actuation system is presented. This end-effector is actuated by a biased SMA wire. The bias force is provided by a 3D-printed custommade spring. The SMA wire is wound over the spring itself. The later one converts the transversal movement of the wire into longitudinal movement along the shaft of the end-effector, which induces the open-close motion of the gripper. The low weight of the overall actuator, along with its wide range rotational displacement capability, makes this actuator a suitable alternative for applications in aerial manipulation with small unmanned aerial vehicles (UAVs). Although all the experimental test presented in this work were performed with a wired communication, the proposed system is equipped for wireless communication using Robot Operating System (ROS). ROS implementation allows for future mobile environment tests, as a flying manipulator on UAVs for example.