Biomimetics and Dextrous Manipulation Lab

DrewBell


  • Email = drewbell@stanford.edu
  • Phone = +1 217-377-7144

Working on: DynamicRotorcraftPerchingMechanisms

SuperSCAMP Design Summary - In Progress

Terminology: 2S, 3S: indicates the number of battery cells in a lithium-polymer(lipo) battery, each cell being nominally 3.7V. A 2S battery is nominally 7.4V and a 3S battery is nominally 11.1V.

Background

The SuperSCAMP project was begun as an extension of the SCAMP project, the Stanford Climbing and Aerial Manuvering Platform, the first robot that is able to fly, perch passively, climb, and take off. Quite impressive in its delicate design, the original SCAMP had severa disadvantages. First, being built upon the bitcraze crazyflie 2.0 platform, it was quite limited in thrust and flight time. The crazyflie has a mass of around 28g, and the added ~10g of mass and even more so the added intertia of the SCAMP system meant the system had a low thrust-to-weight ratio. The thrust-to-weight ratio is a helpful metric in quadrotor design to inform how responsive a quadrotor will be. Partly as a result of old batteries, the flight time of SCAMP was also quite short, down to about 30s per battery charge in the summer of 2017.

Directly as a result of the limited thrust capabilities of the crazyflie, the construction methods use for the climbing mechanism were optimized for extremely light weight, employ carbon fiber rods, kevlar thread, and super glue. The methods of tuning the climbing mechanism included adding superglue (Loctite 404) or removing or softening it with acetone. Several days to weeks were needed to understand the sensitivity of the various adjustment point to tune the mechanism. The critical points were the base of the bow springs, bow string to foot connection. At the base, the joint needed to provide a medium level of torsional stiffness near the end of the most extreme bowing of the spring, and less so as the bow extended out. This stiffness was quite difficult to tune. It also was imporant to get the torsional stiffness around the robot's roll axis current so the leg tracked forward and backward without overlapping or clipping the other leg as they passed each other.

The fact the SCAMP was adept at climbing was in large part due to the careful tuning of it's handlers, but hardly made it a robust system otherwise. The superglue tuning methods that the climbing mechanism could come out of tune simply by the glue drying fully or being on sitting in a box with the foot twisted by contact with the wall.

All this said, SCAMP is an amazing robot with much potential, and so SuperSCAMP was started as a project to make a more robust robot with the same functionality but without the mass and thrust limitations of the original SCAMP.

The goals of SuperSCAMP (SS) are to make a robot that can fly, perch passively on walls, climb, and take off, robustly.

To start with SS, it was desired to add payload capacity to allow for a more tunable climbing mechanism and more thrust for a higher thrust-to-weight ratio, which would permit more aggressive maneuvers. The crazyflie 2.0 platform had brushed DC motors, which at the scale of the craziefly are a good choice for thrust to weight.

The fact the SCAMP was adept at climbing was in large part due to the careful tuning of it's handlers, but hardly made it a robust system otherwise. The superglue tuning methods that the climbing mechanism could come out of tune simply by the glue drying fully or being on sitting in a box with the foot twisted by contact with the wall.

All this said, SCAMP is an amazing robot with much potential, and so SuperSCAMP was started as a project to make a more robust robot with the same functionality but without the mass and thrust limitations of the original SCAMP. Our first test of a bigger quad was a the haktoys quadrotor that Hao Jiang used for some of his perching experiments. Testing (haktoys thrust test) indicated that the quadrotor would only produce 135g of thrust, less than the 150g of thrust that we wanted.

The fact the SCAMP was adept at climbing was in large part due to the careful tuning of it's handlers, but hardly made it a robust system otherwise. The superglue tuning methods that the climbing mechanism could come out of tune simply by the glue drying fully or being on sitting in a box with the foot twisted by contact with the wall.

All this said, SCAMP is an amazing robot with much potential, and so SuperSCAMP was started as a project to make a more robust robot with the same functionality but without the mass and thrust limitations of the original SCAMP. Our first test of a bigger quad was a the haktoys quadrotor that Hao Jiang used for some of his perching experiments. Testing (haktoys thrust test) indicated that the quadrotor would only produce 135g of thrust, less than the 150g of thrust that we wanted.

Next, we about a 200mm quadrotor with 2204 brushless motors, which was terrifyingly powerful and weighed nearly 350g without a climbing mechanism. The benefit of this test, though, was it's introduction to brushless DC motors, which are generally more powerful and efficient per unit mass after a certain level. SuperSCAMP is generally straddling the line between brushed and brushless motors, but we chose to investigate brushless because of their ability to scale with larger robots. The tradeoff of using a brushless motor is the requirement of a electronic speed controller. Whereas brushed DC motors have a mechanical commutator, which acts as a mechanical switch to make the motor keep turning, brushless motors require external circuitry to do this switching. This extra part increase the mass, electrical complexity of the quadrotor, and changes the signal requirements coming out of the flight controller, in this case the crazyflie.

There are advantages to using ESC's, however. First, as opposed to the crazyflie, motors for which are driving in one direction by a low-side drive transitor, some ESC's allow for bi-directional motor drive. It is important to note now, if not discussed more later, that running a general ESC in bidirectional control is using 'senseless' bidirectional control, which means that where is no additional sensor informing the ESC of the motor speed beside the standard back-EMF sensing in the ESC. Most if not all modern ESC's drive two of three motor coils at a time to generate the motor spin, then use use an analog sensor to monitor the voltage induced by the coil spinning near the magnets in the rotor. Sensing this voltage tells the ESC where the motor is rotationally and informs it when to switch to the next coil configuration. In short, this sensing is essential for motor coil switch timing. The challenge of senseless bi-directional control is that when the motor passes through 0-rpm, the back EMF becomes very low, so there is a short period of time when there is (1) no thrust and (2) uncertainty of where the motor is positionally, making it harder for the ESC to drive the motor correctly and avoiding skipping. The sum effect is a short pause in control authority with is enough to cause the vehicle to lose control. In summer of 2017, we did experiment to determine whether bi-directional motor control was useful to graceful perching, recovery from climbing failure, or takeoff, and the result of preliminary testing of each indicated that this period of control loss prevented or contributed to preventing the technique from being useful.

That said, bi-directional control was helpful in one key area of robust perching and recovery. If the robot fell to the ground and landed with rotors facing the ground, bi-directional control was shown to allow the robot to flip itself over to permit flying. Side note: the DYS 1104 motors on 2S was slightly underpowered to do this.

Back to the brushless motors, the 2204 motors (22 means 22mm in diameter, 04 means 4mm in stator height) werer terrifyingly powerful as mentioned before, having on the order of 300g per thrust on each motor. It was abundantly clear that these motors were too large and too heavy. The next iteration, we chose 1104 motors, basically the smallest brushless motors on the market. We learned a few things from these. First, the prop size needs to match both the motor and the craft size. the first iteration had a large body but very small 2" props, which on paper gave enough thrust, but in practice, it was inefficient. We found that 2.5-3" props worked much better. Second, we learned of the idea of KV rating, which is basically a measure of how fast a motor wants to spin. Higher Kv means higher back-emf voltage for a given rotation. More back-EMF means less voltage over the motor coils and less current. But if you put a big prop on a high-kv motor, it won't spin fast enough to generate the back-EMF and will burn a ton of current. A good rule of thumb that I read was to find the lowest KV motor you can find and pair it with the biggest prop (cite). We stated off with 7500kv 1104 motors and quickly found those ran very hot with 3-inch tri-blade props. 3-inch 2-blade bullnose gemfan props ran a little better but were still hot and had short battery life. We next moved to the DYS 1104 5400Kv motors, which were much less hot and had bettery flight time. The pro of high kv rating is more punchiness, i.e. more thrust at the expense of much more current and lower efficiency. Moving to the 5400kv motors meant less thrust but an overall better system, though given the current mechanical build, though still having slightly too little thrust on 2S.

Back to the brushless motors, the 2204 motors (22 means 22mm in diameter, 04 means 4mm in stator height) werer terrifyingly powerful as mentioned before, having on the order of 300g per thrust on each motor. It was abundantly clear that these motors were too large and too heavy. The next iteration, we chose 1104 motors, basically the smallest brushless motors on the market. We learned a few things from these. First, the prop size needs to match both the motor and the craft size. the first iteration had a large body but very small 2" props, which on paper gave enough thrust, but in practice, it was inefficient. We found that 2.5-3" props worked much better. Second, we learned of the idea of KV rating, which is basically a measure of how fast a motor wants to spin. Higher Kv means higher back-emf voltage for a given rotation. More back-EMF means less voltage over the motor coils and less current. But if you put a big prop on a high-kv motor, it won't spin fast enough to generate the back-EMF and will burn a ton of current. A good rule of thumb that I read was to find the lowest KV motor you can find and pair it with the biggest prop (cite). We stated off with 7500kv 1104 motors and quickly found those ran very hot with 3-inch tri-blade props. 3-inch 2-blade bullnose gemfan props ran a little better but were still hot and had short battery life. We next moved to the DYS 1104 5400Kv motors, which were much less hot and had bettery flight time. The pro of high kv rating is more punchiness, i.e. more thrust at the expense of much more current and lower efficiency. Moving to the 5400kv motors meant less thrust but an overall better system, though given the current mechanical build, though still having slightly too little thrust on 2S.

Back to the brushless motors, the 2204 motors (22 means 22mm in diameter, 04 means 4mm in stator height) werer terrifyingly powerful as mentioned before, having on the order of 300g per thrust on each motor. It was abundantly clear that these motors were too large and too heavy. The next iteration, we chose 1104 motors, basically the smallest brushless motors on the market. We learned a few things from these. First, the prop size needs to match both the motor and the craft size. the first iteration had a large body but very small 2" props, which on paper gave enough thrust, but in practice, it was inefficient. We found that 2.5-3" props worked much better. Second, we learned of the idea of KV rating, which is basically a measure of how fast a motor wants to spin. Higher Kv means higher back-emf voltage for a given rotation. More back-EMF means less voltage over the motor coils and less current. But if you put a big prop on a high-kv motor, it won't spin fast enough to generate the back-EMF and will burn a ton of current. A good rule of thumb that I read was to find the lowest KV motor you can find and pair it with the biggest prop (cite). We stated off with 7500kv 1104 motors and quickly found those ran very hot with 3-inch tri-blade props. 3-inch 2-blade bullnose gemfan props ran a little better but were still hot and had short battery life. We next moved to the DYS 1104 5400Kv motors, which were much less hot and had bettery flight time. The pro of high kv rating is more punchiness, i.e. more thrust at the expense of much more current and lower efficiency. Moving to the 5400kv motors meant less thrust but an overall better system, though given the current mechanical build, though still having slightly too little thrust on 2S.

< More documentation coming soon>

Page last modified on May 20, 2018, at 11:34 AM