Smaller, Cheaper, Faster

Increased robot capabilities can be directly linked to the evolution of computer power. The software has been developing over the last 30 plus years and had to wait until computer size (smaller), cost (cheaper) and power (faster) caught up. A human brain is a 100,000 times as large as a guppy’s, but computers will be 100,000 times as powerful as today’s 30 years from now, if computer power continues to double every year or two. In another tack, the first multi-celled animals with nervous systems appeared about 550 million years ago, ones with brains as advanced as guppies’ perhaps 200 million years later. Self-contained robots covered similar ground in about 20 years: at that pace, the remaining 350 million years of our ancestry could be recapitulated robotically in about 35 years.

Chart time line comparing the Evolution of Computer Power and Cost to Equivalent Brain Power

2D to 3D Technology – Increased Computational Power

Dr. Hans Moravec, Seegrid’s Chief Scientist, invented and pioneered the use of evidence grid technology for machine navigation and perception. Evidence grid technology is a software mechanism that breaks the world down into a probabilistically weighted grid. The cells of the grid, called voxels, are filled with statistical evidence about the world and can be used for navigation, object identification and manipulation. Evidence grid technology, properly applied, allows robots to sense and interpret their environment. This approach is widely used and taught in robotics courses throughout the world.

As a PhD student at Stanford’s Artificial Intelligence lab in 1979, Dr. Moravec demonstrated the world’s first vision-guided robot, dubbed the Stanford Cart. Then, in 1984 he invented 2D evidence grids as a mechanism for building maps and navigating using sonar. Twenty-plus years later, 2D evidence grids are the gold standard for autonomous machine navigation. 2D evidence grids have achieved this position because they work, they work simply and elegantly, and they are extensible. He also invented learned sensor models to improve the accuracy and robustness of these evidence grid maps.

Dr. Moravec realized that extending evidence grids to a third dimension would realize an even greater array of capabilities. This work proved both scientifically complex and computationally demanding. In 1992, as computational power started to catch up with his vision, he began experimenting with 3D evidence grids and by 1996 created the world’s first full-fledged 3D evidence grid using stereo cameras.