Collars I used in the past had captive balls within capsules. Each end of the capsule supported an electrical contact. One capsule was sensitive to left/right movements, and the second fore/aft. When they ceased accumulating counts for at least a 24 hour period, the mortality signal would launch. One could just as easily write code to assess movements via GPS integrations, but that might issue a false positive if an animal was denned up and/or a unit was not getting signals.
Has some similar instruments in the past and recovered a couple that had been dropped to the ground. Looking at the data, one could tell that something came along every couple of days and knocked them around for a minute or two.
Similarly many years back a buddy had time lapse film cameras out to assess elk use across some extensive grassy mountain slopes. The elk spotted those and kicked a couple of them 4 to 500 yds down slope.
