Soft sensing

Sensing in soft robotics

Sensing is one of the core element of robotics . Robots need to sense their environment in order to either perform complex tasks or to check for a given task successful execution. For example, for a robot to pick up a box on the floor and to move it to a target location, either the robot and box positions are exactly known, or the robot will need some sort of sensor to navigate to the box. Conversely, even if all positions are supposed known, having way to sense where the box is can improve the capability from the robot to follow instructions. The more a robot is expected to be adaptable, or the more it's environment is unknown, the more sensors are needed.

Soft sensors are also relevant to traditional robotics. Sensors can be used within as well as around robots. One of the particularities of soft sensors is that, although they are obviously suited to be used in soft robots, they can also be used as a soft layer around more traditional robots. Covering conventional robots with sensors is not a new idea [10, 11] but soft sensors are especially relevant even to rigid robots as they can form a soft sensitive skin and wrap the robot, and mitigating possible collisions or insulating the robot from its direct environment for sanitary reasons (pharmaceutical / food industries). 

Soft sensors measure diverse things. It's also important to note that many types of soft sensors exist. Indeed, soft sensors measure the environment in diverse ways and obtain wide ranging information. The soft sensors presented here might often measure strain or pressure, but chemical [27, 37], geo-magnetic [35], or physiological data might also be acquired for example.

Soft sensors design is important. A sensor's overall design is significant, for example although using a single strain gauge (see below) can inform on the body deformation along a single axis, using 4 of them positioned at 90° allows to measure torque. New soft sensing technologies often start by taking the shape of a waveguide [24] as it one of the most straightforward designs [26], but can evolve to more complex designs to either improve the quality [21] or the quantity of data [25] collected. 

Soft resistive sensors

One of the most straightforward soft sensing technologies consists of making soft sensors in which deformation changes internal electrical resistance. Resistance is basically how much the sensor opposes electrical current flow [4]. Two main avenues have been explored to do so. 

The first one consist on creating a well defined soft conductive path enclosed in a soft material. When the conductive path's geometry changes, its resistance changes and can be interpreted as strain. Although this strategy is directly inspired by traditional strain gauges made using metal thin films deposited on polymers [5], in order to make these sensors soft,  the materials used have to be modified. A number of sensors have used soft elastomers as a encapsulant for traces made of conductive liquids such as liquid metals [2,6,7] or room temperature ionic liquids [3,8,9]. A consequence of using such materials is that the sensors are capable of very large deformations. 

The second option is to directly imbue a soft elastomer with conductive particles. These might be metal [13], carbon black [12] particles and recently conductive liquid droplets have also been used [14]. In these composites, characterized as piezoresistive [15], electrons can "jump" from one conductive particle to another. This makes them loosely conductive, at the price of increased mass and stiffness that their non-doped counterparts.  When mechanically stressed, compressed or stretched for example, the inter-particles distance changes and with it the composite's resistance. 

Soft capacitive sensors

These sensors are quite popular in soft robotics. Indeed they are fairly easy to manufacture as they are composed of two soft conductive electrodes separated by a soft dielectric material [16]. 

One of the main advantages of capacitive sensing is that, because it is being used by traditional industries,  specialized Application Specific Integrated Circuits (ASICs [18]) have been developed for this particular purpose. One of the best exemple is you cell phone, in which multi-touch detection is nowadays performed using capacitive sensing elements [19]. 

Contrary to traditional industry where rigid materials are used, in soft robotics capacitance sensing is often based on material deformation. Capacitance obeys the following equation C = εrε0A/d where C is the capacitance, A the electrode area, d the distance between the electrodes and εrε0 respectively relative permitivity of the material between the electrodes, and ε0 the vacuum permeability. In soft robotics, the material deformation oftentimes leads to a changed of distance (d) between top  and bottom electrodes which turns into a change in the measured capacitance.

Soft optical sensors

Optical fibers are commonly used both to transmit data and as strain sensors. Glass, however, is famously brittle. When optical fibers are used as sensors they can only measure (although very precisely) up to a few strain percentage points [20]. As the astute reader will have no doubt understood by now, deformations in soft robots are much larger. As a consequence, once again soft robots require new optical materials to measure large strains. 

Several slightly different approaches have been proposed [21,22,23], however, in general soft materials are usually fairly poor light waveguides. As a consequence, soft optical sensors usually rely on measuring the amount of light escaping the waveguide to deduct the sensor's deformation. One of the drawback of using optical loss is that these sensors usually can't be more than a few tens of centimeters long at most.

Soft barometric and acoustic sensors

Barometric sensors [29] based soft sensors have been developed as well. Manufactured using rigid materials such as silicon wafers [30], these highly specialized micro-electromechanical systems  (MEMS) [28] can measure ambient fluid pressure while keeping with a small form factor, allowing them to be encapsulated in soft elastomers [31]. Pressure applied on the elastomer pad is transmitted and detected by the barometer. 

Barometric sensors capture fairly slow changes in pressure. Acoustic sensors also capture pressure waves, although at a much higher frequency. As a consequence, another way to detect contacts and deformations is to actively generate and measure acoustic waves within soft robots [32]. One approach consist in generating acoustic waves of a wide range of frequencies [33]. The acoustic waves propagate within the soft robot's structure and are eventually reflected back to an acoustic sensor.  The soft robot's acoustic signature is then compared to the soft robot know acoustic signature. The difference between the original signature and the measured one can inform on the soft robot position and deformation or even be used to recognize the possible materials in contact. 

A slightly different approach consist in using a well defined acoustic signal. Indeed, in a similar manner to previously presented resistive and optical soft sensors, single frequency acoustic waves can also be guided in soft structures, albeit with many similar drawbacks such as due to the use of a fairly lossy medium [34].

Soft magnetic sensors

Recently soft sensors based on electromagnetic fields have also been demonstrated. Such sensors have been manufactured by mixing ferromagnetic powders into soft elastomers and using traditional (but small) magnetometers to measure the surrounding magnetic field. Essentially, this is akin to making soft magnets, and measuring how the magnetic field changes as the magnet is deformed [1]. Some soft and thin magnetic sensors are considered for electronic skin [36] applications [35].