Apply Dielectric Electro-Active Polymer Sensor on an Autonomous Robot

The aim of this project was to design, fabricate and test smart sensors based on Dielectric Electro-Active Polymer (DEAP). The sensors was used on HExapod Cognitive auTonmously Operating Robot (HECTOR) as a test platform for the EMICAB project. The EMICAB project takes a holistic approach to robotics, where nature and especially insects are a source inspiration. Part of the EMICAB project was the development of smart sensors and the integration of them into and a distributed sensor network with a redundancy of sensors mimicking sensory organ functions found on limbs of for instance stick insects.
These Bio-inspired sensors will provide sensory feedback information to the robot both about state of the robot and its surroundings.

Figure 1.1: Rendered design model of the six-legged robot HECTOR [1]

Project background

The EMICAB project takes a holistic approach to the engineering of articial cognitive systems. The project`s purpose is to integrate smart body mechanism in intelligent planning and control behavior of the motors [2]. The robot will be used as a testbed for various biologically inspired control and navigation approaches. It has 18 rotatory drivers which constitute the joints in each of the six legs. These drives are fully self-contained, including a brushless DC motor, a harmonic drive gear, power electronics, sensors and electronics to control the drive independently of a central computer [2]. Sensors will be placed on the robot and transmit to the bus the exact position of the joint.

Figure 1.2 Ventral view of the right rear leg of a typical insect, showing thelocations of the main types of sense organs. On the left; Organs in or on the cuticle (exoskeleton): campaniform sensilla, hair plates, spines, and sensory hairs. On thr right; The arrangement of organs within the leg: chordotonal organ, strandreceptor, stretch receptor, and muscle receptor organ. Not all of these senseorgans are present together in every species of insect [3].

Insects typically move around by walking, flying or occasionally swimming. When moving on the ground, many insects colopt a tripodal joint in which they walk with there legs touching the ground in alternative triangles, allowing a rapid get stable movement. In order to walk the insect relies on a sensor network where mechanoreceptors are on part of this. Mechanoreceptors can be subdivided into three categories: proprioceptors, tactile receptors and stress receptors. Proprioceptors are sense organs that encode information of the position and the movements of parts of the body relative to one another. In insects, leg sense organs that serve this function are chordotonal organs, strand receptors, stretch receptors, muscle receptor organs and hair plates. Tactile receptors are those that encode information about contact with external objects. Spines and sensory hairs belong to this group. Stress receptors provide information about forces acting on the exoskeleton. Campaniform sensilla ful-fill this role in the insect leg. The boundaries between functional groups are not always sharply de ned [3]. In the following project the chordotonal organ and the stretch receptors will be used as the main idea for the sensor design.

Figure 1.3: The femoral chordotonal organ in a locust. The organ in this insect is anchored to the cuticle at several sites, as well as to the tibia and to the tendon of the tibial flexor muscle (muscles not shown). The physical arrangement may be different in the organs in other legs or other insects. Most of the sense cells of the chordotonal organ are stretched during flexion of the tibia whereas others are stretched during extension. [3]. Typically, a chordotonal organ consists of a mass of as many as 100 sensory cells that are attached to one or more strands of connective tissue. The connective tissue is stretched between the cuticle and a movable point, either the tendon of a muscle or a part of the next leg segment, as shown in figure 1.2. When the strand of connective tissue is stretched during leg movement, the branches of the sensory cells that are embedded in it are also stretched, resulting in the activation of the receptors. Each of the individual sensory cells in a chordotonal organ sends a separate sensory signal back to the central nervous system. Chordotonal organs are distributed roughly one to a leg segment, although it is possible for a segment to have more than one. In insect legs, most of the sensory cells of chordotonal organs are activated by flexion of the more distal segment of the leg. That is, the chordotonal organ in the femur is activated by flexion of the tibia, and the chordotonal organ in the coxa is activated by flexion of the femur [3].

Problem formulation

The core of the problem is to design and fabricate smart sensors which can be placed on the leg of HECTOR and measure the joint position. In this project different design of sensors was tested to full-fill this requirement. Furthermore a testing platform should be designed and built in order to test both sensor material, sensor design and contact material.

Figure 1.4: Requirements for the movement of the joint between the tibia and the femur, the other two joints found on the leg have similar movement requirements.

The EMICAB consortium have listed a set of requirements which the joint sensors should full-fill:

• ˆThe accuracy should be 1 degree or below.
• ˆThe sensors should be able to do the full mount of the joint which is 120°.
• ˆThe maximum speed of the joint 57 [mm/s], so the sensor should be tested up to this speed.
• ˆThe readout electronics has to be fast enough to get values from the sensor, so at full speed it is still has to achieved a 1 degree resolution which means with at least 120 Hz.

Outcome

The outcome of this project is to apply DEAP joint sensor on the limb of Hector. The idea is to mount the sensor as a ligament found in a human knee. The design idea can be seen figure 1.3 where the tendon of tibia extensor and tendon of tibia flexor muscle is working in parallel. When the insect moves the leg while it is walking, extensor muscle is stretching and flexor muscle is Figure 1.1: Rendered design model of the six-legged robot HECTOR [1]. As seen in figure 1.5 similar design was made. Here the sensors end with the electrical connections is fixed on the non-moving part of the limb. The other part of the sensors are fixed on the moving part of the joint and follows it during rotation. In this configuration it is needed to minimize the high friction between the silicone surface of the sensor and the motor cover of the joint. This has be done with a surface coating for instance by mounting the sensor in a highly stretchable fabric. Furthermore, the repeatability will be tested of the sensor in the full movement range and develop embedded electronics for capacitance measurements which t inside the limb of HECTOR.

Figure 1.5: The idea for mounting the sensor on the leg of the robot. The end of the sensors with the electrical connection are fixed on the non-moving part of the limb. The other part of the sensors are fixed on the moving part of the joint and follow it during rotation.

Conclusion

In this project I was using the DEAP material from Danfoss PolyPower A/S, as active material in smart sensors for the EMICAB hexapod robot. During this project different sensor designs were tested in order to optimize the way the sensor was mounted on the joint of the robot. The idea was to design bio-inspired from insect’s legs. Two designs were tested, a sheet stretch sensor and a rolled-up sensor which would optimize the active sensor area.

Finally, the DEAP sensor was fabricated, coated with silicone in order to protect the sensor from scratches and damages. An additional sleeve of textile polyester was applied to the sensor to reduce the friction between the silicone layer and the plastic of the joint. During this initial phase of sensor building a basic understanding the DEAP material and its characteristics were obtained. Sensors based on the DEAP material were tested with different contact materials to the silver electrodes to and the most appropriate contact material. Conductive tape which has a high electrical conductivity and its easy handling. A challenge was found in this project was to maintain the properties of this unique DEAP material when contact were applied to it.

The electronics was made in order to convert the capacitance in to frequency by using a 555 timer circuit. The frequency was converted into a voltage by using LM2917 frequency to voltage converter. The electronics which was designed to full-fill the accuracy requirements stated.

Moreover a precise manually manipulated test platform, here called the optical rail was used to test the sensors in a linear stretch. The sensitivity of the linear stretch for the 110 mm long sensor was calculated 6 [mV/deg] and for 130 mm long sensor it was 4 [mV/deg]. Afterwards the sensors was mounted on the joint and was tested dynamically were the material was bend over the joint. Due to the motors not being available, a servo motor was mounted on the limb. Arduino was used to control servo both speed ans position of the limb. The speed of the motor was measured and it was found out that the maximum speed of the full joint is 54.5 [ms/deg]. The sensors was tested dynamically by using Labview program and saved the data in a le. On the joint both a 110 mm and 130 mm long sensor were mounted. The 130 mm long sensor was able to measure between 20 and 120 degrees and the 110 mm long sensor was able to measure between 0 and 55 degrees, so the two sensors completed the full movement of the 120 degrees stated in the requirements. The sensor for both cases had the same linear behavior. The slopes which was found for linear stretch and dynamical stretch were the same. The sensitivity for dynamical stretch for the 110 mm long sensor was calculated 10 [mV/deg] and for 130 mm long sensor it was 8 [mV/deg]. Therefore the microcontroller which will be used for the bus has to be able to measure at least 10 [mV/deg] accuracy.

It can also concluded that each sensor needs to be calibrated in order to know the slope. The reason behind this is that the sensors were made by hand and there were differences between the fabricated sensors, because of small ripples when then silicone were applied on the sensor, also when the sensor cut out or a conductive tape contacted to the sensor.

As a final conclusion the fabricated sensors were good enough to full-fill the requirements stated in the problem formulation.

Bibliography

[1] Rendered design model of six-legged robot HECTOR http://www.gizmag.com/insect-inspired-hector-walking robot/18421/picture/133185/
[2] EMICAB(Embodied Motion Intelligence for Cognitive Autonomous Robots): http://ndeaa.jpl.nasa.gov/nasa-nde/newsltr/WW-EAP_Newsletter13-1.pdf
[3] Fred Delcomyn, Mark E. Nelson and Jan H. Cocatre-Zilgien. The International Journal of Robotics Research 1996 15: 113 DOI: 10.1177/027836499601500201