Multihop Paths and Key Predistribution in Sensor Networks Guy Rozen Contents Terminoligy (quick review) Alternate grid types and metrics k-hop coverage ◦ Calculation ◦ How to optimize Complete two-hop coverage Terminology DD(m) – Distinct distribution set of m points DD(m,r) – DD(m) with maximal Euclidian distance of r DD*(m)/ DD*(m,r) – DD(m)/ DD(m,r) on a hexagonal grid DD(m,r) – Denotes use of the Manhattan metric DD*(m,r) – Denotes use of the Hexagonal metric Ck(D) – Maximal value of a k-hop coverage for some DDS D Scheme 1: Let D v1 , v2 ,..., vm be a distinct difference configuration. Allocate keys to notes as follows: ◦ Label each node with its position in 2 . ◦ For every ‘shift’ u generate a key ku and assign it to the notes labeled by u vi , for i 1, 2, ..., m . 2 Alternate grid types and metrics In a regular square grid, where the plane is tiled with unit squares, sensor coordinates are in fact 2. 1,1 0,0 Alternate grid types and metrics In a regular square grid, where the plane is tiled with unit squares, sensor coordinates are in fact 2. 1,1 0,0 In a hexagonal grid, where the plane is tiled with hexagons, seonsor coordinates can be depicted as 1, 0 1 2, 3 2 | , 1 2, 3 2 1 32 0,0 12 Moving between grid types y 2y The linear bijection : x, y x , transitions from a 3 3 hexagonal grid to a square one. 3 Alternatively, can be seen as doing , , 2 2 Moving between grid types y 2y The linear bijection : x, y x , transitions from a 3 3 hexagonal grid to a square one. 3 Alternatively, can be seen as doing , , 2 2 Theorem 1: If D v1 , v2 ,..., vm is a DD* m , then D v1 , v2 ,..., vm is a DD m . Similarly, if D is a DD m then 1 is a DD* m . Moving between grid types y 2y The linear bijection : x, y x , transitions from a 3 3 hexagonal grid to a square one. 3 Alternatively, can be seen as doing , , 2 2 Theorem 1: If D v1 , v2 ,..., vm is a DD* m , then D v1 , v2 ,..., vm is a DD m . Similarly, if D is a DD m then 1 is a DD* m . Proof: Since is linear: vi v j vk vl vi v j vk vl Moving between grid types It is important to note that does not preserve distances. Theorem 2: If D is a DD* m, r then D is a DD m, r 2 . If D is a DD m, r then 1 D is a DD* m, r 3 2 . Alternate metrics Manhattan/Lee metric: The distance between two points i1 , j1 and i2 , j2 is i1 i2 j1 j2 . For example, a sphere of radius 2: Theorem 3: For r , a DD m, r is a DD m, r and a DD m, r is a DD m, 2r Alternate metrics Hexagonal metric: The distance between two points is the amount of hexagons on the shortest path between the points. For example, a sphere of radius 2: Theorem 4: For r , a DD* m, r is a DD* m, r and a 2 DD* m, r is a DD* m, r 3 k-Hop Coverage Definition: Ck D is the amount of vectors of the form v l i 1 i vi when i , i 1, 2,..., m ; i i ;0 l k k-Hop Coverage Definition: Ck D is the amount of vectors of the form v l i 1 i vi when i , i 1, 2,..., m ; i i ;0 l k Theorem 5: Suppose that D is used in Scheme 1. Then the k-hop coverage of the scheme is equal to Ck D k-Hop Coverage Definition: Ck D is the amount of vectors of the form v l i 1 i vi when i , i 1, 2,..., m ; i i ;0 l k Theorem 5: Suppose that D is used in Scheme 1. Then the k-hop coverage of the scheme is equal to Ck D Proof: When using Scheme 1, we know that a pair of nodes sharing a key are located at vi u , v j u , hence the vector vi v j is both a difference vector of D and a one hop path when using Scheme 1. Hence, an l-hop path between paths is composed of difference vectors from D. k-Hop Coverage Theorem 6: Let D be a DD* m and let D be a DD m such that D D . Then the k-hop coverage of D is equal to the k-hop coverage of D . Proof: is a bilinear bijection. Maximal k-hop coverage First, we define a set of integer m-tuples: m m m H k a1 , a2 ,..., am | ai 0, ai k i 1 i:ai 0 Maximal k-hop coverage First, we define a set of integer m-tuples: m m m H k a1 , a2 ,..., am | ai 0, ai k i 1 i:ai 0 Simply put, the some of elements in each tuple is zero and the sum of positive elements is k. Some examples for m=3: 0, 0, 0 H 0 ; 1, 1, 0 H1 ; 2, 2, 0 , 2, 1, 1 H 2 Maximal k-hop coverage First, we define a set of integer m-tuples: m m m H k a1 , a2 ,..., am | ai 0, ai k i 1 i:ai 0 Simply put, the some of elements in each tuple is zero and the sum of positive elements is k. Some examples for m=3: 0, 0, 0 H 0 ; 1, 1, 0 H1 ; 2, 2, 0 , 2, 1, 1 H 2 Lemma 7: i w H k1 , z H k2 w z H k3 where 0 k3 k1 k2 ii w H k , z H k , w z w z H k where 1 k3 k1 k2 iii Any element of H k may be written as the sum of k1 elements 1 2 3 1 from H1 . Maximal k-hop coverage Theorem 8: The k-hop coverage of a DD m is at most k H i 1 m if and only if all the vectors a v i 1 are distinct. i i i , with equality with a1 , a2 ,..., am k Hj j 0 Maximal k-hop coverage Theorem 8: The k-hop coverage of a DD m is at most k H i 1 m if and only if all the vectors a v i 1 i i i , with equality with a1 , a2 ,..., am k Hj j 0 are distinct. m Proof: The difference vectors in D are all of the form ai vi i 1 with a1 ,..., am H1 . Hence: V is a sum of at most k difference vectors k m V ai vi | a1 ,..., am H j j 0 i 1 Maximal k-hop coverage Proof (cont.): We can now say: k k k m Ck D 1 ai vi | a1 ,..., am H j H i 1 H i i 1 j 0 i 1 i 0 Maximal k-hop coverage Proof (cont.): We can now say: k k k m Ck D 1 ai vi | a1 ,..., am H j H i 1 H i i 1 j 0 i 1 i 0 Corollary 9: The two-hop coverage of a DD m is at most: 1 m m 1 m 2 m 3 m m 1 m 2 2m m 1 4 1 m m 1 m 2 m 6 4 Maximal k-hop coverage Proof: We proved that the two-hop coverage is at most H1 H 2 . H1 contains m m 1 elements. As for H 2 we have this lovely table: Maximal k-hop coverage - bounds We would like to show that Theorem 8’s bound is tight. i Naïve approach: vi 2k 1 , 0 , i 1, 2,..., m Maximal k-hop coverage - bounds We would like to show that Theorem 8’s bound is tight. i Naïve approach: vi 2k 1 , 0 , i 1, 2,..., m Lemma 10: The k-hop coverage of a DD m given by D v1 , v2 ,..., vm meets the bound of Theorem 8 m c v i 1 i i 0 for all c 2k Hi i 1 Maximal k-hop coverage - bounds We would like to show that Theorem 8’s bound is tight. i Naïve approach: vi 2k 1 , 0 , i 1, 2,..., m Lemma 10: The k-hop coverage of a DD m given by D v1 , v2 ,..., vm meets the bound of Theorem 8 m c v i i i 1 0 for all c 2k Hi i 1 Proof: 2k m m m Bound not met ai vi bi vi ci vi 0; c1 ,..., cm H i i 1 i 1 i 1 i 1 m Assume ci vi 0 when c1 ,..., cm H l for some l 1, 2,..., 2k i 1 vector c is a b where a H l , b H l 2 2 Bound not met Bh Sequences Definition 1: A - an abelian group. D v1 , v2 ,...vm A - a sequence of A's elements. D is a Bh sequence over A if all sums vi1 vi2 ... vih with 1 i1 ... ih m are distinct. Elements may be used more than once. Bh Sequences and DDC Theorem 11: Let k 2 be a fixed integer and D v1 , v2 ,..., vm 2 . Then D is a DD m with maximal k-hop coverage D is a B2 k over 2 Bh Sequences and DDC Theorem 11: Let k 2 be a fixed integer and D v1 , v2 ,..., vm 2 . Then D is a DD m with maximal k-hop coverage D is a B2 k over Proof: Assume D is a B2 k over 2 2 i If vi v j for i j we get a contradiction vi v j ii If vi v j vi v j for i j , i j we get a contradiction vi v j vi v j or i i and j j iii If D does not have maximal k-hop coverage by Lemma 10 there is a c 2k m H i so that i 1 c v i i i 1 0 a is c's positives m m i 1 i 1 and b a - c 2k t v1 ai vi 2k t v1 bi vi we get a contradiction D has maximal k-hop coverage Bh Sequences and DDC Proof (cont.): Assume D is a DD m with maximal k-hop coverage If D is not a B2 k sequence then there are two sums as seen in the Bh definition which are equal m m m m a v b v ;1 a , b i 1 i i i 1 i i c a b H t gives i 1 i i 1 m c v i 1 i i i t 2k 0 we get a contradiction. Using Bh sequences to build a DDC Construction 1: Let k 2 be a fixed integer. Let q be a prime power so that q 2 k 1 ab where a, b are coprime. Then there exists a set of dots X 2 which is doubly periodic with periods a, b and that for each rectangle R of size a b, R X is a DD q with maximal k-hop coverage. Using Bh sequences to build a DDC Construction 1: Let k 2 be a fixed integer. Let q be a prime power so that q 2 k 1 ab where a, b are coprime. Then there exists a set of dots X 2 which is doubly periodic with periods a, b and that for each rectangle R of size a b, R X is a DD q with maximal k-hop coverage. Proof: Example 1 gave a B2 k sequence over there is an isomorphism sequence v1 , v2 ,..., vq q 2 k 1 a a q 2 k 1 b with q elements. Also, We get a B2 k b Define x, y x mod a, y mod b , so we have : and can define X as v 2 2 a | v v1 , v2 ,..., vq Every a b rectangle R will give us X R v1 , v2 ,..., vq , a B2 k sequence b Maximal k-hop coverage - bounds Theorem 12: Let k 2 be a fixed integer, and c 16 21 k . Then there exists a DD m, r with maximal k-hop coverage such that m cr1 k . Maximal k-hop coverage - bounds Theorem 12: Let k 2 be a fixed integer, and c 16 21 k . Then there exists a DD m, r with maximal k-hop coverage such that m cr1 k . Proof: S 2 set of points in a r / 2 radius from the origin. Note that S 4 r 2 O r . Let q be the smallest prime power for which q k 2r and we get q 2r 1k We now define: q k 1 when q is even k a q 1 2 when q k 3mod 4 k k 1mod 4 q hen w 2 1 q b q 2 k 1 a Maximal k-hop coverage - bounds Proof (cont.): As we saw in Construction 1, a a b rectangle contains q dots. A shift T of S contains S q ab dots at average T so that T X S q ab . We define D T X where D m S q ab 4 r 2 q 2r 16 21 k r 1 k D is contained in a sphere of radius r 2 , and in a a b rectangle 2 which is a DD q coverage D is a DD m, r with maximal k-hop coverage. Maximal k-hop coverage - bounds Proof (cont.): As we saw in Construction 1, a a b rectangle contains q dots. A shift T of S contains S q ab dots at average T so that T X S q ab . We define D T X where D m S q ab 4 r 2 q 2r 16 21 k r 1 k D is contained in a sphere of radius r 2 , and in a a b rectangle 2 which is a DD q coverage D is a DD m, r with maximal k-hop coverage. Corollary 13: 1 2k Using theorem 2, 6, and 12 we can say that for c 16 21 k 2 3 there exists a DD* m, r with maximal k-hop coverage so that m cr1 k Maximal k-hop coverage - bounds What is the minimal r so that a DD m, r with maximal k-hop coverage is possible? We denote this r as r k , m . Maximal k-hop coverage - bounds What is the minimal r so that a DD m, r with maximal k-hop coverage is possible? We denote this r as r k , m . Theorem 14: For an integer k 2, k mk 1 16 o mk r k , m mk o mk 2 k ! k Maximal k-hop coverage - bounds What is the minimal r so that a DD m, r with maximal k-hop coverage is possible? We denote this r as r k , m . Theorem 14: For an integer k 2, k mk 1 16 o mk r k , m mk o mk 2 k ! k Proof: (Upper bound proven in Theorem 12) k Remember Ck D H i . i 1 Define B a1 , a2 ,..., am H k : i : ai 0 2k . We get B m! . 2 m 2k ! k ! k So we can say i 1 m! m2k 2k 2k Hi o m o m k !2 m 2k ! k !2 Maximal k-hop coverage - bounds Proof (cont.): Vectors in Ck D are at most k difference vectors. All difference vectors are r at most. All vectors Ck D are inside a kr sphere, which contains at most kr O r vectors. 2 m2k 2 We get Ck D 2 o m 2 k kr O r . k! Maximal k-hop coverage - bounds Proof (cont.): Vectors in Ck D are at most k difference vectors. All difference vectors are r at most. All vectors Ck D are inside a kr sphere, which contains at most kr O r vectors. 2 m2k 2 We get Ck D 2 o m 2 k kr O r . k! For a hexagonal grid we present an equivalent term r * k , m . Theorem 15: For an integer k 2, k 2 mk 2 1 16 k o mk r * k , m m o mk 3 k ! k 3 2 Proof: Theorem 2 & 14. Maximal k-hop coverage - bounds We will give special attention to the case k=1. Theorem 16: 2 2 m o m r 1, m m o m , where 0.914769 Maximal k-hop coverage - bounds We will give special attention to the case k=1. Theorem 16: 2 2 m o m r 1, m m o m , where 0.914769 Proof: 2 For DD m, r we have m 2 DD m, r : m 2 r O r 2 3 Lower bound r o r Upper bound 2 Maximal k-hop coverage - bounds We will give special attention to the case k=1. Theorem 16: 2 2 m o m r 1, m m o m , where 0.914769 Proof: 2 For DD m, r we have m 2 DD m, r : m 2 r O r 2 3 Lower bound r o r Upper bound 2 Theorem 17: 231 4 m o m r 1, m * 21 2 31 4 m o m Proof: Analogous hexagonal result from [2]. Maximal k-hop coverage - bounds Finally, using results in [2] we can prove: Theorem 19: r 1, m 2 m o m 2 3 m o m r * 1, m 2 m o m , 1.58887 Minimal k-hop coverage What is the smallest value for a k-hop coverage? Minimal k-hop coverage What is the smallest value for a k-hop coverage? Theorem 20: The k-hop coverage of a DD m is at least km m 1 Minimal k-hop coverage What is the smallest value for a k-hop coverage? Theorem 20: The k-hop coverage of a DD m is at least km m 1 Proof: The 1-hop coverage of a DD m is m m 1 , the amount of difference vectors. Define S1 the set of initial difference vectors. We pick u= d , e S1 with maximal d and if need be, e . We can assume d 0, e 0 and can say that S1 is composed of: S1 x, y | x, y S1 , x 0 or x 0 and y 0 , S1 x, y | x, y S1 For i 1: Si w i 1 u | w S1 w i 1 u | w S1 Si S j for i j , Si m m 1 . Minimal k-hop coverage Lemma 21: For an integer k 2, suppose D is a DD m where two differences are not parallel. Then the k-hop coverage of D is more than km m 1 . Minimal k-hop coverage Lemma 21: For an integer k 2, suppose D is a DD m where two differences are not parallel. Then the k-hop coverage of D is more than km m 1 . Proof: Define u and Si as in Theorem 20. Let v be a difference vector not parallel to u , and the most perpendicular. The k-hop coverage of D is at least S1 S 2 ... S k 2v We saw that Si m m 1 , and we can see that 2v S1 ... S k , since p w i 1 u p w p v 2 p v p 2v . Minimal k-hop coverage Lemma 21: For an integer k 2, suppose D is a DD m where two differences are not parallel. Then the k-hop coverage of D is more than km m 1 . Proof: Define u and Si as in Theorem 20. Let v be a difference vector not parallel to u , and the most perpendicular. The k-hop coverage of D is at least S1 S 2 ... S k 2v We saw that Si m m 1 , and we can see that 2v S1 ... S k , since p w i 1 u p w p v 2 p v p 2v . Lemma 21 can be used to prove Theorem 21: For an integer k 2, suppose D is a DD m . D meets the bound of Theorem 20 if and only if it is equivalent to a perfect Golomb ruler. Complete 2-hop coverage For a prime p 5 , we will show a construction of a DD m with complete 2-hop coverage. That ensures a two-hop path between a point x and any other grid point within a 2 p 3 2 p 1 rectangle centered at x. Height Width Complete 2-hop coverage For a prime p 5 , we will show a construction of a DD m with complete 2-hop coverage. That ensures a two-hop path between a point x and any other grid point within a 2 p 3 2 p 1 rectangle centered at x. Height Width Definition 2 (Welch Periodic Array): Let be a primitive root modulo a prime p. Define the Welch periodic array: R p i, j 2 | j i mod p. R p is double periodic: i, j R p i p, j p 1 R p for , . Equivalent points: A i, j i , j A i i p, j j p 1 . j Example of an array 10 9 8 6 5 4 3 2 1 1 3 7 p5 2 3 4 5 6 7 8 9 10 11 i Complete 2-hop coverage Lemma 23: Let d and e be integers such that d 0 mod p and e 0 mod p 1 . Suppose that R p contains dots A i1 , j2 , B i1 d , j2 e as well as A i2 , j2 , B i2 d , j2 e . Then A A, B B . Complete 2-hop coverage Lemma 23: Let d and e be integers such that d 0 mod p and e 0 mod p 1 . Suppose that R p contains dots A i1 , j2 , B i1 d , j2 e as well as A i2 , j2 , B i2 d , j2 e . Then A A, B B . Proof: i1 j1 mod p j2 i2 mod p j1 j2 e 1 0 mod p j1 e i1 d mod p i2 d j2 e mod p Since e 0 mod p 1 we get j1 j2 mod p 1 and from that i1 i2 mod p Complete 2-hop coverage From Lemma 23 we conclude: if R p contains i, j , i d , j then d 0 mod p. if R p contains i, j , i, j e then e 0 mod p 1 . A vector d , e can occur at most once as a difference between two points in R p , within a particular p 1 p rectangle. Complete 2-hop coverage We now define a DD m by using dots from R p . Complete 2-hop coverage We now define a DD m by using dots from R p . Construction 2: p an odd prime, i, j 2 is such that i, j , i 1, j 1 R p j ◦ Do such points exist? Why yes! i 1 1 mod p and also j 1 1 1 1 1 i 1 mod p Complete 2-hop coverage We now define a DD m by using dots from R p . Construction 2: p an odd prime, i, j 2 is such that i, j , i 1, j 1 R p j ◦ Do such points exist? Why yes! i 1 1 mod p and also j 1 1 1 1 1 i 1 mod p S is a p 1 p rectangle bounded by i, j , i p 1, j , i, j p 2 and i p 1, j p 2 . There are p 1 dots in S . ◦ Why p 1 dots? Remember what is! Complete 2-hop coverage We now define a DD m by using dots from R p . Construction 2: p an odd prime, i, j 2 is such that i, j , i 1, j 1 R p j ◦ Do such points exist? Why yes! i 1 1 mod p and also j 1 1 1 1 1 i 1 mod p S is a p 1 p rectangle bounded by i, j , i p 1, j , i, j p 2 and i p 1, j p 2 . There are p 1 dots in S . ◦ Why p 1 dots? Remember what is! Since R p is periodic, it also contains dots at i, j p 1 , i p, j and i p 1, j p . Our, actual construction is a configuration B , which is S and the above dots. Meet j p B j p 1 A p2 p 3 S Central region j 1 B j A A i i 1 i p i p 1 - Vital statistics Contained in a p 1 p 2 square. Has a border region of width 2 which contains exactly 5 points. Has a central region which is a p 3 p 2 rectangle. The central region contains p 3 dots. One column is empty. A A A and B B and there are no other equivalent points. j Example of 10 9 8 6 B’ A’ 5 4 3 2 B 1 A 1 3 7 p5 2 3 A’’ 4 5 6 7 8 9 10 11 i Complete 2-hop coverage Lemma 24: The configuration B is a DD p 2 , with all points in a p 1 p 2 rectangle. Complete 2-hop coverage Lemma 24: The configuration B is a DD p 2 , with all points in a p 1 p 2 rectangle. Proof: We have already shown that B contains the dots as needed. Suppose that X and Y and X and Y are distinct dot pairs with the same difference vector d , e . Suppose d 0, p, p and/or e 0, p 1 , p 1 . It is impossible for one of X , Y , X , Y to be in the central region. Why? This is why j p B j p 1 A p2 p2 p 3 j 1 B j p 1 A i i 1 A i p i p 1 Complete 2-hop coverage Proof (cont.): X , Y , X , Y must be outside the central region, but no two pairs have the same difference vector. So d 0, p, p and e 0, p 1 , p 1 and since all dots are in a p 1 p 2 rectangle d 0 mod p and e 0 mod p 1 . Lemma 23 X X , Y Y . X X Y Y point pairs are not distinct X X X , X and Y , Y must be on configuration border, but we've already seen that leads to a contradiction, and so we have our proof. Complete 2-hop coverage Motivational boost: In order to show a 2 p 3 2 p 1 rectangle with complete 2-hop coverage, we will show that any vector d , e where d p 1 and e p 2 can be expressed as a difference vector of B.. Complete 2-hop coverage Motivational boost: In order to show a 2 p 3 2 p 1 rectangle with complete 2-hop coverage, we will show that any vector d , e where d p 1 and e p 2 can be expressed as a difference vector of B.. Lemma 25: Any vector of the form d , e , where 0 d p 1 and 0 e p 2 where d , e are integers is a sum of two difference vectors from B. Complete 2-hop coverage Motivational boost: In order to show a 2 p 3 2 p 1 rectangle with complete 2-hop coverage, we will show that any vector d , e where d p 1 and e p 2 can be expressed as a difference vector of B.. Lemma 25: Any vector of the form d , e , where 0 d p 1 and 0 e p 2 where d , e are integers is a sum of two difference vectors from B. Proof: Remember the p 1 p rectangle S from Construction 2. Define A as a restriction of R p to the 2 p 2 2 p sub-array whose lower left corner coincides with S . illustrated 3 4 1 2 S illustrated D1 S B. Since R p is doubly periodic with S 's size, the points in D1 occur in the other quadrents as well. 3 4 1 2 Complete 2-hop coverage Proof (cont.): We want to show that any vector d , e as defined is a difference of two points in A . We assume d 0. Assume e 0 and can say that there is a i , j D1 so that d mod p e 1 It is easy to see that i , j , i d , j e R p , and since d , e 0 and j i are small enough, i , j , i d , j e A . In the case where e 0 we can just pick i , j D3 and follow use same concept. Complete 2-hop coverage Proof (cont.): Now that any vector d , e is a difference of two points in A . All we need to show is that every two points in A can be connected using two difference vectors from B. You'll see this more easily if I show you. D1 to D1 (or any Dx to Dx) By definition, all such vectors are a single difference vector from B. 3 4 1 2 D1 to D3 (or D2 to D4) We use the vector 0, p 1 to cross quadrents, then use another difference vector from B to reach our destination (this second one exists since R p is doubly periodic. 3 4 1 2 D1 to D3 (or D2 to D4) We use the vector p, 0 to cross quadrents, then use another difference vector from B to reach our destination (this second one exists since R p is doubly periodic. 3 4 1 2 D1 to D4 We use the vector p, p 1 to cross quadrents, then use another difference vector from B to reach our destination (this second one exists since R p is doubly periodic. 3 4 1 2 D3 to D2 We use the vector p, p 1 to cross quadrents, then use another difference vector from B to reach our destination (this second one exists since R p is doubly periodic. 3 4 1 2 Complete 2-hop coverage Lemma 26: For an integer t 3, let F be a set of integers where: a b c d e F t 1 F t 1 , t 2 ,..., 1 1, 2,..., t 1 t 1 1, t 1 , t 1 F i F \ 1, t 1 , t 1 with i 0 if i 0 and i F \ 1, t 1 , t 1 then i t F Then for each positive integer where 1 t 1 there are i, j F so that j 1. Complete 2-hop coverage Lemma 26: For an integer t 3, let F be a set of integers where: a b c d e F t 1 F t 1 , t 2 ,..., 1 1, 2,..., t 1 t 1 1, t 1 , t 1 F i F \ 1, t 1 , t 1 with i 0 if i 0 and i F \ 1, t 1 , t 1 then i t F Then for each positive integer where 1 t 1 there are i, j F so that j 1. Why do we need this? Complete 2-hop coverage Lemma 26 motivation: Let us look at all vectors of the form 1, y where y 0 present in B. We will show how the right-sided coordinates of these vectors satisfy the conditions of Lemma 26 for t p 1. Complete 2-hop coverage difference vectors Lemma 26 motivation: Let us look at all vectors of the form 1, y where y 0 present in B. We will show how the right-sided coordinates of these vectors satisfy the conditions of Lemma 26 for t p 1. a B has p difference vectors as defined above. Proof of (a) j p B j p 1 A p2 p 3 One of the p 2 columns is empty, so we have p 3 vectors j 1 B j A A i i 1 i p i p 1 Complete 2-hop coverage Lemma 26 motivation: Let us look at all vectors of the form 1, y where y 0 present in B. We will show how the right-sided coordinates of these vectors satisfy the conditions of Lemma 26 for t p 1. a B has p difference vectors as defined above. b Except for 1, p , all of B 's 1, y vectors satisfy y p 2. Proof of (b) j p B j p 1 A p2 y p 1, p 3 but this y p vector is not legitimate j 1 B j A A i i 1 i p i p 1 Complete 2-hop coverage Lemma 26 motivation: Let us look at all vectors of the form 1, y where y 0 present in B. We will show how the right-sided coordinates of these vectors satisfy the conditions of Lemma 26 for t p 1. a B has p difference vectors as defined above. b Except for 1, p , all of B 's 1, y vectors satisfy y p 2. c The vectors 1,1 , 1, p 2 , 1, p are all in B. Proof of (c) j p B j p 1 A 1, p 2 p2 p 3 1, p j 1 B j A i 1,1 i 1 A i p i p 1 Complete 2-hop coverage Lemma 26 motivation: Let us look at all vectors of the form 1, y where y 0 present in B. We will show how the right-sided coordinates of these vectors satisfy the conditions of Lemma 26 for t p 1. a b c d B has p difference vectors as defined above. Except for 1, p , all of B 's 1, y vectors satisfy y p 2. The vectors 1,1 , 1, p 2 , 1, p are all in B. B 's 1, y vectors can't all satisfy y 0. Proof of (d) – case one j p B j p 1 A p2 p 3 1, y y0 j 1 B j A A i i 1 i p i p 1 Proof of (d) – case two j p B j p 1 A p2 j 1 B j A i 1,1 i 1 No dots p 3 1,1 A i p i p 1 Complete 2-hop coverage Lemma 26 motivation: Let us look at all vectors of the form 1, y where y 0 present in B. We will show how the right-sided coordinates of these vectors satisfy the conditions of Lemma 26 for t p 1. a b c d e B has p difference vectors as defined above. Except for 1, p , all of B 's 1, y vectors satisfy y p 2. The vectors 1,1 , 1, p 2 , 1, p are all in B. B 's 1, y vectors can't all satisfy y 0. If 1, y where y 1, p is a vector in B , 1, y p 1 is not. Complete 2-hop coverage Lemma 26 motivation: Let us look at all vectors of the form 1, y where y 0 present in B. We will show how the right-sided coordinates of these vectors satisfy the conditions of Lemma 26 for t p 1. a b c d e B has p difference vectors as defined above. Except for 1, p , all of B 's 1, y vectors satisfy y p 2. The vectors 1,1 , 1, p 2 , 1, p are all in B. B 's 1, y vectors can't all satisfy y 0. If 1, y where y 1, p is a vector in B , 1, y p 1 is not. e is proven thusly: If 1, y p 1 were in B then by Lemma 23 the points involved would be equivalent to those making the vector 1, y . Such points would be on the border region, and that is impossible. Complete 2-hop coverage Lemma 26 motivation: By similar means, we can show Lemma 26 works for vectors of the form x,1 when t p. We will now face insurmountable suspense… Complete 2-hop coverage Proof (of Lemma 26): F \ 1, t 1 , t 1 contains t 2 elements F must contain precisely one element of each pair i, i t for i 2,3,..., t 1. Suppose, for a contradiction, t -1 that cannot be expressed as a difference of two elements from F . Suppose 1 and then: 1, t 1 F 1 , t 1 F However, since 1 t 1 t one of them is in F contradiction Complete 2-hop coverage Proof (of Lemma 26, cont.): Suppose 1 F does not contain a pair of integers which differ by 1 If t is odd then F \ t 1 contains at most t 1 2 positive integers and at most t 1 2 negative integers F contains at most t 1 1 t integers, which contradicts a . If t is even then in order for F to be of size t 1, F \ t 1 must contain t 2 positive integers (which must be odd) and t 2 negative integers (also odd). This implies that for each odd integer 1 i t we have i, i t F and this contradicts e . Complete 2-hop coverage Theorem 27: Let p 5 be a prime. The distinct difference configuration B achieves complete two-hop coverage on a 2 p 3 2 p 1 rectangle, relative to the central point of the rectangle. Complete 2-hop coverage Theorem 27: Let p 5 be a prime. The distinct difference configuration B achieves complete two-hop coverage on a 2 p 3 2 p 1 rectangle, relative to the central point of the rectangle. Proof: We notice that all vectors from the central point d , e are such that 0 d p 1, 0 e p 2. When d 0, e 0, Lemma 25 proves that the vector is composed of two of B ' s difference vectors. When either d 0 or e 0 we can use Lemma 26. For 0 e p 2, we have shown that there are two vectors in B which are 1, y , 1, y so that y y e and so 0, e 1, y 1, y . Similarly for d , 0 where 0 d p 1. Conclusion and open problems We have shown maximal k-hop coverage as B2 k over 2 . We used a construction of B2 k over to produce a DD m, r with maximal k-hop coverage and of the order of r1 k dots. We have found a bound for r k , m (verifying the order above). Could we find tighter bounds? What is the exact value for small k and m? The questions above also hold for the hexagonal grid and the alternate metrics. We have constructed a DD m, r with complete 2-hop coverage from the center of a rectangular region. The rectangle’s region is of order m2. Can we find a construction for significantly larger rectangles? For circles? Can we find constructions for k-hop coverage where k>2?