Ultrasonic sensors are super common in industry. Modules vary by FOV, distance, update rate, mounting, size, etc. We did a semi-formal comparison of the more po

Ultrasonic sensors are super common in industry. Modules vary by FOV, distance, update rate, mounting, size, etc. We did a semi-formal comparison of the more popular mainland Chinese units last year. They all have the problem where partial blocks to their FOV (eg. thin post or foliage) or non-perpendicular surfaces generate wildly fluctuating output. Other problem scenarios include dirtied sensor lenses, glass and liquids (feel like reversing through a fish tank?). While these issues are basically inherent to the physics, approaches like sensor arrays, post-processing or visual/inertial sensor fusion, inset mounting and washdown can help to resolve it. The good thing is at the low end they are dirt cheap, in the order of $0.50 or less at volume, with typical nominal ranges of 45 or 60cm, though 4m may be readily achieved in practice. So for an array of three units with dedicated post-processing you may be talking an anticipated final production cost of around $2 per unit. That said, your control plane is still going to need to remain seriously skeptical of output, as something as simple as mud can totally destroy your readings. One may reasonably suspect that Tesla made a strategic decision to move away from ultrasonic owing to the reliability issues pervasive to the technology.