C.2 Source Annotation

Dealing with Irregular Terrain

The most common mechanical gait is moving on wheels. However, as robots move off the road, wheels can often hinder mobility rather than aid it. Cars rarely react well to uneven surfaces and often get stuck in softer surfaces such as mud or sand. Animals however, have been traversing harsh terrains for millions of years. From kangaroos hopping across the Australian continent to human, legs have dealt with many show-stoppers for wheels such as stairs, and recently, mud, slush, ice and snow. Robots, however, are quickly gaining off-road capability. One day, there might be a giant hexapodal vehicle walking down Locust Walk, avoiding students and plowing through the occasional snow or construction. To sum, [1] claims that:

"Humans maneuver easily over uneven terrain. To maintain smooth and efficient gait the motor system needs to adapt the locomotor output to the walking environment."

Apart from the rather science fiction use of versatile off-road robots as vehicles, robots that can deal with irregular terrain can be used in various ways. [2] suggests "urban search and rescue, that require both small size and the ability to locomote through highly rubbled terrain." Robots can be sent to navigate dangerous conditions under collapsed buildings to find trapped individuals. Additionally, if the robots can run on rough terrain, they can also interact with it, allowing for dangerous logging or mining to be done by robots without the need to risk human life. There are numerous hazardous conditions where robots can be used instead of humans.

Existing Technology

There are many existing robots designed for irregular terrain that take a variety of forms and leg number. [3] presents

"…free gaits for structural symmetrical four-legged robot capable of performing statically stable, omnidirectional walking on irregular terrain."

Another design is a multi-leg centipede model proposed by [4] is also designed for uneven terrain.

"…By creating contact points of the first legs adequately on the environment, the robot can climb over obstacles and be navigated successfully."

There is a six legged waker called the DLR Crawler which is based on the fingers of a hand robot that has been adapted to locomotion. [5] developed a reactive walking algorithm which

"…allows the robot to autonomously master uneven terrain and obstacles with height differences within the normal walking height…For the second gait algorithm inspired bu Cruse's rules, which were identified for forward walking stick insects, an implementation has been found for the DLR Crawler that fives the robot full omnidirectional mobility.."

Other approaches include analyzing established models of wheel/terrain interactions and develop a model for terrain discrimination to produce appropriate gait. This also include active sensing functions for the robots [2].

"…Active sensing the terrain allows self-adaptation of a robot's gait to improve locomotion efficiency. The gait bounce approach to terrain classification has been shown to work for a small class of terrains."

Of course, there is also the RHex which uses six actuators, simple clock driven terrain independent system [6] which has

"…significant 'intrinsic mobility'-the traversal of rugged, broken, and obstacle ridden ground…RHex achieves fast and robust forward locomotion traveling at speeds up to one body length per second…"

Potential Biological Solution

Anticipatory changes in movement pattern could aid in navigating in rough terrain. When a person approaches an icy surface, she braces herself and changes her gait to gliding motions to avoid uncontrolled slip. When that person comes to a rocky surface, she aims to walk on the flatter smooth rocks.

[7] states that most animals

"…can make moment-to-moment adjustments to motor programs and also plan new paths by anticipating interactions between their own movements and potential obstacles. Some compensation for unpredictable terrain can be achieved by mechanical means. A running cockroach for example exploits the inertial and damping forces of its body and limbs to clamber over uneven surfaces … However, in most situations sensory feedback is an important component of movement adaptation."

The current feasibility of this solution is limited because of the computation required to match biological terrain categorization and adaptation.

Alternatively there could be more compliance, robustness optimization which is suggested by [8] which is naturally accomplished by the musculoskeletal system.

Value of Sources

The keywords "terrain" "sensing" "gait" were entered into Google Scholar and Scopus. While many hits on Google Scholar were on topic, most were dated. Not one of the top five searches were in the last 6 years, one dating back to 1983. The Scopus search results returned [7] which presents an interesting insect model, the caterpillar, to study changes in motor behavior due to sensory information. This was a potential candidate for the focal biology paper of this review but upon source evaluation, I decided to choose another paper. The journal of publication, Journal of Comparative Physiology A has a ISI JCR Impact factor of 1.852. Comparing this to the ISI JCR impact factors for other journals in physiology, this is not a high impact factor. The highest impact factor was Physiol Rev which had an impact factor of 37.726. However, this value quickly reduces to single digits slightly down the list. Even so, by impact factor, J Comp Physiol A is ranked 46th and is not a high impact journal. Both authors have h indices of 2, co-authorships in the teens, and single digit publications according to Scopus. While this is an interesting subject and they use an interesting model to explore motor anticipation, these factors make the paper less impressive. Scanning articles from the results found in Scopus, keyword searches were modified to different combinations of "gait" or "gait analysis", "irregular OR uneven" and "terrain OR surface", which yielded many papers relevant to my interest and applied to searches in Scopus and INSPEC and Pubmed.

Open Problems

Some problems that exist are with processing problems with sensing and the speed of reaction. Even in caterpillars [7] the learned obstacle only affects the later segments while the initial segments stumble with the obstacle.

Speed is one of the competing issue in robotics. While slow thought out actions may be appropriate for some circumstances, generally, long time delays are not viewed positively. Instead, the cockroach's reliance on the elastic properties of its body instead of relying on undue processing makes it faster than more complex organisms such as humans in movement.

Also, the primary robotics paper [8] presents an issue of robustness against perturbation, saving power during locomotion.

"…The most efficient leg configuration in walking occurs when the leg is not compressed during stance…however, such a stiff leg behavior results in extensive forces at touchdown and liftoff, which increases the risk of damage. By contrast, compliant legs in walking systems enable the reduction of impact forces but may require more energy."

While [8] provides a model to optimize the previous point, there has yet to be a physical device implementing the suggested behavior.

Annotations

Robotics Paper

Paper [8] uses a bipedal spring-mass model to find the balance between robustness and efficiency of different gaits. These two factors were approached from the point of view of leg stiffness and movement patterning. Different terrain were considered and optimized solutions were found for each. Unexpected external perturbations were also considered. With these concepts, power expenditure could be minimized without sacrificing performance and could result in longer battery life that allows for increased functionality for long usages.

Since this paper was not considered in C1, the source validation was repeated and the results are as follows. Bioinspir Biomim is the 6th ranked journal in robotics using ISI JCR sorting by impact factor. It has an impact factor of 1.367. Thus, this seems like a reasonably source. The most senior author, Andre Seyfarth is from a university, University of Jena, the Lauflabor Locomotion Laboratory, in Germany. He has an h index of 11 and 32 publications according to Scopus, 34 co-authorships, which is a respectable amount. For subject area, he is very broad but some of his subject areas cover computer science, engineering, and physics which would give this paper credibility. However, he is also listed for Social Sciences and Psychology which are rather unrelated to the field of interest here. When googled, he showed up as the head of the Locomotion Lab at the Unverisity of Jena. The lab's stated research areas were the Biomechanics of Long Jump, Leg Design and Control, Control of Human Locomotion, and Bionics which are all directly relevant and appeases the subject area concern.

Precursors

This paper has a number of highly cited references. One was [9] which has been cited 850 times in Scopus and 1473 times in Google Scholar. This highly cited paper from 1990 presented an idea that bipedal machines can be made where walking is the natural dynamic mode and analyzed the physics of the walking cycle. Another reference was [10] which was cited 395 times in Scopus and cited 580 times in Google Scholar. This paper was published in a high impact multidisciplinary paper, Science, and presented some of the first robots based on passive-dynamic walkers to produce low-powered robots that have remarkably human walking motions.

Successors

This paper has not yet been cited by any sources in Scopus. This could be a consequence of its recency (it was published in December 2010). It has been cited once in Google Scholar. This was a conference proceeding [11] from the same first and last author as the main Robotics Paper.

Biology Paper

The paper [1] analyzed 19 healthy human volunteers walking over a hydraulically actuated platform where the platform would accelerate downwards on random trials. The different muscle and joint (specifically the ankle) responses, including muscle activity and response time, were analyzed. They conclude that

"…force feedback from ankle extensors increase the locomotor output through positive feedback in late stance. In midstance the effect of force feedback was not observed…"

Since this paper was not analyzed in C1, I do so here. There are two senior authors from two different institutes. The first is Jens Beyer Nielsen who, according to Scopus has 262 documents, 4104 citations, an h index of 31, and has over 150 co-authors. His area of expertise is in medicine, neuroscience as well as other sciences which is related to the paper's topic of concern. The second senior author was Thomas Sinkjaer, who, again according to Scopus, has 195 documents, 2203 citations, an h index of 26, over 150 co-authors, and also specializes in medicine, neuroscience, amongst other topics. Additionally, it was published in J Neurophysiol with an ISI JCR impact factor of 3.483 and 22 in the impact factor physiology rankings which makes it a good journal though not the most impactual.

Precursors

The most cited paper in the references [12] was sited 288 times according to Scopus and 313 times according to Google Scholar. This 1997 paper concerns the neurological basis of stepping and draws a connection between peripheral sensory information and the human lumbosacral spinal cord in its modulaiton of efferent output which may help generate stepping. A second reference [13] explores how load is sensed and translated into altered gait and posture.

Successors

This article has been cited 7 times in Scopus. Most of these articles were about human locomotor muscle. However, one article [14] was about a new design of "robotic lower limb exoskeleton for gait rehabilitation…" and cite the paper as a characterization of normal human walk as well as to characterize the soleus response.

Wikipedia Stub

The most similar to my topic, terrain sensing is the Wikipedia article on terrestrial locomotion. This article mentions some animal gaits that deal with terrestrial terrain.
Terrestrial Locomotion

References

1. R. af Klint, J.B. nielsen, T. Sinkjaer, M.J. Grey "Sudden Drop in Ground Support Produces Force-Related Unload Response in Human Overground Walking", J Neurophysiol, vol. 101, pp. 1705-1712, 2009.
2. A.C. Larson, R.M. Voyles, J. Bae, et al. "Evolving Gaits for Increased Discriminability in Terrain Classification", Proceedings of the 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems.
3. C. Huai, Y. Fang, K. Zhang "Gait Design and Stable Control for a Symmetrical Four-legged Robot on Irregular Terrain", Proceedings of the 7th World Congress on Intelligent Control and Automation.
4. S. Inagaki, T. Niwa, T. Suzuki "Follow-the-Contact-Point Gait Control of Centipede-Like Multi-Legged Robot to Navigate and Walk on Uneven Terrain", The 2010 IEEE/RSJ International Conferences on Intelligent Robots and Systems.
5. M. Gorner, T. Wimbock, G. Hirzinger "The DLR Crawler: evalutaion of gaits and control of an actively compliant six-legged walking robot", Industrial Robot: An International Journal, vol. 36, no. 4, pp. 344-351, 2009.
6. U. Saranli, M. Buehler, D.E. Koditschek "RHex: A simple and highly mobile hexapod robot", International Journal of Robotics Research, vol.20, issue 7, pp. 616-631, 2001.
7. L.I. van Griethuijsen, B.A. Trimmer "Caterpillar crawlin gover irregular terrain: anticipation and local sensing", J Comp Physiol A, vol. 196, pp. 397-406, 2010.
8. J. Rummel, Y. Blum, A. Seyfarth "Robust and efficient walking with spring-like legs", Bioinsp. Biomim., vol. 5, no. 046004, pp. 1-13, 2010.
9. R. McGeer "Passive dynamic walking", International Journal of Robotics Research, vol. 9, issue 2, pp. 62-82, 1990.
10. S. Collins, A. Ruina, R. Tedrake, M. Wisse "Efficient bipedal robots based on passive-dynamic walkers", Science, vol. 307, Issue 5712, pp.1082-1085, 2005.
11. J. Rummel, A. Seyfarth "Passive stabilization of the trunk in walking", Proceedings of SIMPAR 2010 Workshops, pp. 127-136, 2010.
12. S.J. Harkema, S.L. Hurley, U.K. Patel, et al. "Human lumbosacral spinal cord interprets loading during stepping", J Neurophysiol, vol. 77, issue 2, pp.797-811, 1997.
13. J. Duysens, F. Clarac, H. Cruse "Load-regulating mechanisms in gait and posture: Comparative aspects", Physiol Rev, vol. 80, issue1, pp. 83-133, 2000.
14. P.C. Kao, C.L. Lewis, D.P. Ferris "Short-term locomotor adaptation ot a robotic ankle exoskeleton does not alter soleus Hoffmann reflex amplitude", J NeuroEngineeirng and Rehabilitation, vol. 7, issue 1, article 33, 2010.