Woah that’s a little pessimistic don’t you think. Yes we are working with a form of NiTiNOL and yes it does come with a lot of drawbacks. But we have been hard at work to try and bring this technology down to earth. We still have a long way to go but we have already made plenty of improvements to existing actuators.

Do you have any experience with NiTiNOL? We would love to know what we can do to make our stuff better. The typical application of our current muscles is in a one-way mechanism where the passive load extends the muscle back to a relaxed state. We have also tested 2-muscle systems where a second muscle causes the first to relax, creating two way motion.

​@@deltaroboticsinc My apologies for sounding too pessimistic and all, I'm trying to be realistic because I really went through a lot of articles facing the same problems of efficiency, speed, cost and precision control. This is because anthropomorphic robotics is a highly complex subject that is still an incredible challenge for the brightest minds of our time, *but it doesn't mean that you should give up, it means that you should be critical of simple solutions presented to such complex problem.* And being honest, I only know about soft actuators through scientific articles, I'm not certified in science and engineering of materials. And I would suggest you to seek such professional when proceeding with your project. However, from an outsider perspective, I'm critical of certain aspects of your actuator in this regard. Even though you can achieve some kind of motion control with opposite/antagonistic Nitinol muscles, it comes always with the cost of efficiency, since you are wasting energy by making the muscles fight each other for a required position. But you could use a nitinol (or other type of thermally activated actuator) with sections that could be individually activated for a better motion control, there are other ways you could achive that and I added it to my new comment below this one. You should take into consideration the speed of actuation of organic muscles (40cm/s), the mechanical disadvantage they are while attached to the bones (10:1) and the speed and force at the tip of the human limbs (4m/s at running speeds). A system of any type would be too heavy if you positioned them at the same mechanical disadvantage of the human muscles, so you would need to find compromises with the position and power output of these actuators.

​@@deltaroboticsinc Also, I tried to post links for articles on shape memory actuators and the like, so I could actually be helpful for your endeavor, but youtube automatically deletes comments with too many links on it. So, I will just post the name of the article and the relevant paragraph in question: "High-Performance Robotic Muscles from Conductive Nylon Sewing Thread" I can't copy paste the image here in this comment, but they show a couple of graphs showing the change in time between the contraction and relaxation of the artificial muscles made out of nylon. However, these don't work similarly to Nitinol, since these aren't shape memory polymers. Nylon and Polyethylene are a type of thermal expansion actuators, since they expand horizontally and shorten vertically when heated, simulating a muscle contraction. Plus, they actually show an interesting method for position control by using changing voltages and active cooling using cooler fans. This type of thermal expansion actuators can be used to make artificial muscle with insane amounts of force generation, for example, this silicone rubber reinforced with twisted carbon fiber threads can lift 12,600 times its won weight. The only problem is that it needs the carbon fibers to be connected to both ends so the expansion of the rubber can force the fibers apart. "Theory of the tensile actuation of fiber reinforced coiled muscles" I say that because a possible way of achieving motion control with thermal actuators is by individually heating and/or cooling specific sections of the artificial muscle. "It is well known that human muscles in maximum contraction exert a large force in the line of the tendon, often exceeding 1000 lb., and that this force is dependent on the initial length of the muscle as shown by Evans & Hill (1914), Reijs (1921), and others. In everyday life these forces are not noticed, for the muscles work at a mechanical disadvantage, usually of about 10: 1, which gives a high gear-ratio and allows rapid movement. The precise mechanical disadvantage is dependent on the angle of the acting joint and this finally determines the force which can be exerted." Source: "THE EFFECT OF LIMB POSITION IN SEATED SUBJECTS ON THEIR ABILITY TO UTILIZE THE MAXIMUM CONTRACTILE FORCE OF THE LIMB MUSCLES" "Figure 4. Force-velocity relationship. Data were obtained from Figure 1 in Ralston et al. (1949) using specialized software (ImageJ 1.51q8, NIH, United States). This modified version represents the force-velocity relationship of the in vivo human pectoralis major muscle." Source: "On the Shape of the Force-Velocity Relationship in Skeletal Muscles: The Linear, the Hyperbolic, and the Double-Hyperbolic" (Aso, this source also corrects me, since I thought the limit for the speed of linear muscular contraction in the human body was 40cm/s, it can even reach 120cm/s. Since the mechanical disadvantage can vary from body part from body part, this could even reach 12m/s of tip speed in a limb) "Although FAMs are typically used pneumatically, this research uses them hydraulically. Both pneumatic and hydraulic artificial muscles have their own advantages and disadvantages. However, the use of hydraulic McKibben muscles is more appropriate for an efficiency study with application to bipedal walking robots. Pneumatic actuation offers the convenience of venting air to the environment and also maximizes the actuator’s compliance2. However, due to the compressibility of air, valve-controlled pneumatic systems typically operate below 30% efficiency3; and also have a slow response time and poor position control authority4. On the other hand, valve-controlled hydraulic systems have been found to operate near 60% efficiency3. This is due to very little energy being lost while compressing the working fluid, given that hydraulic fluid has a much higher bulk modulus than air. Due to hydraulic fluid’s higher bulk modulus, hydraulic actuation offers a faster response time and better position control authority. For these reasons, hydraulic McKibben muscles will be studied." Source: "Efficiency testing of hydraulic artificial muscles with variable recruitment using a linear dynamometer"

​@@excell211Hey, thanks a bunch! We appreciate you sharing those papers. We've dug into most of them already, and the new ones are going straight onto our reading list. You're right, we're still learning as we go - no experts here! We don't anticipate our initial prototypes matching the power of electric motors. Our first muscle design may have a lengthy cool-down period, but that's where awesome folks like you come in. We need your honest feedback to get these things right. We'd be honored to have you join our Discord server: discord.gg/ByGPpvn7 SMAs offer impressive recovery strain potential. Even achieving 1% of their 20,000 PSI maximum translates to a powerful 200 PSI actuator. While NiTiNOL is widespread, we're aware of superior (yet more expensive) alternatives. We're also researching 3D printable shape memory polymers - a potential revolution in the field. We recognize the challenges of thermo-mechanical systems, but believe sparking interest can boost demand, lower costs, and ultimately spur the development of better solutions. Our vision is a future where biomimetic robots are the norm. We aim to challenge the perception of actuators as rigid components, introducing the concept of soft, flexible systems. SMAs are a promising starting point on this journey.

Right on the money, we will need cooling at some point and are actively working on a few different ideas to make that happen.

Hopefully the tech can be refined on larger-scale hobbyist projects first and then optimized to use as little power as possible. This is still a very early version of this tech that won't make it to prosthetics until its researched more. Luckily, the control system is open source so that might be possible!