Table 2. Quantitative and qualitative properties of selected studies on activity and intention recognition.
| Reference | F.latent.infty | F.plan.synth | F.duration | F.action.sel | F.probability | F.struct.state | F.non.monoton | F.complexity | Method | Scenario | N.states | N.plan.length | N.classes | N.subjects | M.accuracy | M.conf.based | |
| 1 | [1] [4] | ▪ | ▪ | □ | ▪ | ▪ | ▪ | ▪ | 1 | BD | M | 70,000† | 20 | 3 | 23 | □ | □ |
| 2 | [2] [56] | ▪ | ▪ | □† | ▪ | ▪ | ▪ | ▪† | 1 | BD | OM | – | – | ◊ | sim | □ | □ |
| 3 | [3]– | ▪ | ▪ | ▪ | ▪ | ▪ | ▪ | ▪ | 1 | BPF | O | 70,000† | 15† | 10 | 6 | ▪ | ▪ |
| 4 | [4] [20] | ▪ | ▪ | □ | ▪ | ▪ | ▪ | □ | t | BPl | K | 10,000† | – | 3 | sim | □ | ▪ |
| 5 | [5] [20] | ▪ | ▪ | □ | ▪ | ▪ | ▪ | ▪ | 1 | BP | K | 70,000 | 6 | 5 | sim | ▪ | ▪ |
| 6 | [57] [19] | ▪ | □ | □ | □ | ▪ | ▪ | □ | 1 | BD | A | 200,000 | 5† | 6 | 6 | □ | ▪ |
| 7 | [12] [19] | ▪ | □ | □ | ▪ | ▪ | ▪ | □ | 1 | BD | K | 70,000 | 40 | ◊ | 2 | □ | □ |
| 8 | [21] [18] | □ | ▪ | □ | □ | ▪ | ▪ | ▪ | 1 | BD | O | 250,000† | – | 5 | 5 | ▪ | □ |
| 9 | [58] [59] | □ | ▪ | □ | □ | ▪ | ▪ | ▪ | t | NBN | M | 1,000† | – | 15† | sim | □ | ▪ |
| 10 | [60]– | □ | ▪ | ▪ | □ | ▪ | □ | ▪ | 1 | BH | K | 28 | 6 | 6 | – | ▪ | ▪ |
| 11 | [61]– | □ | ▪ | □ | □ | ▪ | □ | ▪ | 1 | BH | A | 300† | 12† | 15 | 3 | ▪ | □ |
| 12 | [62] [19] | □ | ▪ | ▪ | □ | ▪ | □ | ▪ | 1 | BRP | K | 96 | – | 13 | 2 | □ | ▪ |
| 13 | [63] [19] | □ | ▪ | ▪ | □ | ▪ | □ | ▪ | 1 | BRP | O | 3,500† | 3 | 3 | 2† | □ | □ |
| 14 | [64] [56] | □ | ▪ | □ | □ | ▪ | ▪ | ▪ | t | OML | M | – | 20† | 4 | 14 | ▪ | ▪ |
| 15 | [29] [19] | □ | ▪ | ▪ | □ | ▪ | ▪ | ▪ | 1 | BD | AK | 528† | – | 33 | 3† | ▪ | ▪ |
| 16 | [65] [19] | □ | □ | ▪ | □ | ▪ | □ | □ | 1 | NMH | O | 720† | – | 2 | 1 | □ | ▪ |
| 17 | [66] [19] | □ | □ | □ | □ | □ | □ | □ | 1 | LDL | K | – | 15 | 6 | sim | □ | ▪ |
| 18 | [67] [19] | □ | □ | □ | □ | □ | □ | □ | 1 | LDL | AK | – | 24† | 8 | 3 | ▪ | □ |
| 19 | [7]– | □ | □ | □ | □ | □ | □ | □ | t 2 | OG | M | – | 50† | ◊ | 2† | □ | □ |
| 20 | [68] [19] | □ | □ | □ | □ | ▪ | □ | □ | 1 | LP | A | 100† | 40† | 7† | 6 | ▪ | □ |
| 21 | [8] [18] | □ | □ | ▪ | □ | ▪ | □ | □ | 1 | BMF | A | 20,000 | 14† | 14 | 3 | ▪ | ▪ |
“▪” = feature included in study.
“□” = feature not included.
“x †” = value/property x not explicitly stated in study description.
“–” = value unknown.
“◊” = property not meaningful considering target of study.
Method codes: L: logic-based (DL = description logic, P = combined with possibility theory). B: using some variant of sequential Bayesian filtering (exact: H = HMM or extension, D = other DBN, Pl = transformation into a planning problem, P = partially observable Markov decision process; approximate: PF = particle filter, RP = Rao-Blackwellized particle filter, MF = marginal filter). N = Non-sequential Bayesian inference (MH = Metropolis-Hastings, BN = unrolled Bayes Net). O = other exact method (G = some kind of grammar, ML = Markov Logic net). Scenario codes: K = kitchen task, A = other activities of daily living, O = office, M = miscellaneous other scenario.
We consider the first five studies as CSSM-like approaches.